Then, the deep dive this week is how 4.7 + two other Claude releases - Claude Routines and Claude Managed Agents - creates magic.

You’re not going to want to miss this one.

Arize: Close the Eval Gap

Everyone’s building AI products. Almost nobody is evaluating them.

Arize is the platform that closes that gap. Their open source tool Phoenix has 2M+ monthly downloads. Their enterprise platform Arize AX is used by Uber, Booking.com, Duolingo, and hundreds more.

One command in Claude Code and your agent is fully instrumented. You can see every trace, every decision point, every failure in a clean dashboard before a single user hits it.

Try Arize

P.S. You can get a full year of Arize’s pro plan for free ($1260 value) with my bundle.

There’s a billion AI news articles every week. Here’s what actually mattered.

The Week’s Top News: Anthropic Shipped a Weaker Model on Purpose

Anthropic released Claude Opus 4.7 on Thursday. It’s the best publicly available model for agentic coding. They also put the model they decided was too dangerous to release to the public in the same comparison table and showed you the gap on a chart.

Mythos Preview is gated to defensive security teams and roughly 40 companies. If you’re reading this as a builder, that rightmost column is not your reality. It’s the ceiling, not the floor. Everything else in this piece is available to you today.

Quick decoder for the table below: SWE-bench measures whether an AI can fix real GitHub issues autonomously. Terminal-Bench tests command-line coding tasks. Humanity’s Last Exam is a set of questions designed to be too hard for any AI. The percentages are success rates - higher is better.

This is the most unusual launch chart in frontier AI history.

Look at the rightmost column. Mythos Preview beats Opus 4.7 by 13 points on SWE-bench Pro, 10 on Humanity’s Last Exam, 13 on Terminal-Bench. Per Anthropic’s Mythos Preview system card, the model developed working exploits for patched Firefox vulnerabilities 181 times in the CyberGym eval, compared to 2 for Opus 4.6. Anthropic decided those capabilities were too dangerous to ship, gated Mythos to defensive security teams only, then put it in the comparison table of the model they will sell you.

That’s not how any other lab operates. OpenAI makes the released model look like the frontier. Google does the same. You’re never supposed to know what they’re sitting on. Anthropic just told you exactly what they held back, why they held it back, and showed you the gap on a chart.

The gap between those with the most powerful models and those without continues to expand.

The Other News That Mattered

• Perplexity rebuilt Mint by adding full Plaid bank integration inside Perplexity Computer. Mint died in 2024 because nobody would pay for budgeting as a standalone product. Perplexity already has the $20/month subscription. Finance is just a retention feature now, not a business model. That’s the structural reason AI companies keep eating vertical SaaS.

• Shopify gave every AI coding agent direct write access to 5.6 million store backends through the new AI Toolkit. Products, orders, inventory, SEO, images. One prompt now does what used to cost $200 to $500 a month in apps plus a $2,000 SEO audit. Shopify isn’t building the agent. They’re building the protocol that makes every agent a Shopify agent. That’s the platform play.

• Claude Code shipped /ultrareview this week alongside Opus 4.7. When you run it, Claude spins up multiple AI reviewers in the cloud that check your code in parallel. Every bug one reviewer flags gets independently confirmed by another before it shows up in your results. No more reviewing the reviewer.

• Google launched Skills in Chrome on desktop this week. You save a prompt once, hit forward slash, and it runs on any page you’re viewing. Macro automation for your browser, no code required. It’s the kind of feature that sounds small until you realize you’ve been retyping the same prompts into Gemini every single day.

• Windsurf 2.0 launched with Devin baked directly into the IDE. You plan locally, hand tasks to Devin with one click, close your laptop, and Devin opens PRs while you sleep. OpenAI offered $3B for Windsurf. Google paid $2.4B for the CEO and top researchers. Cognition picked up the rest: the IDE, 350 enterprise customers, and $82M ARR, for a reported ~$250M - roughly 3x ARR for a developer tool with enterprise traction (Cognition didn’t disclose terms). That deal is looking better every week.

Resources

• I think you need to start plugin maxxing. Zedchi went through 101 official Claude Code plugins and picked the 36 actually worth installing across 8 categories.

• Dwarkesh released one of the best AI podcasts yet in his combative interview with Jensen Huang. The highlight for me was Dwarkesh pointing out the holes in Jensen’s publicly stated statements. Jensen spends the whole podcast trying to sell his chips. Impressive performance for the 25 year-old Dwarkesh.

Tools

• Google Skills in Chrome lets you save any prompt and run it on any page with one click. Macro automation for your browser, no code required. Rolling out to all desktop users now.

• VoxCPM2 just made studio-quality voice cloning free. 30 languages, 48kHz output, runs on a gaming laptop. ElevenLabs charges $0.16 per minute. VoxCPM2 charges your electricity bill. I wrote about what this means for ElevenLabs’ $500M raise.

Subscribe now

Claude Routines and Claude Managed Agents: The Builder’s Guide

Claude Opus-4.7 made all the headlines. But the real magic is in combining Opus-4.7 with two features Anthropic released earlier in the week: Routines and Managed Agents.

Routines automate your own work. Managed Agents automate your product’s work.

That’s today’s deep dive:

1. Routines vs Agents

2. What this means even if you never write code

3. Claude Routines: what they actually are and builds for every role

4. Claude Managed Agents: the production layer and 3 real deployments

5. What’s live, what’s preview, and what breaks

Subscribe now

1. Routines vs Agents

Before this, building a production AI agent meant two jobs.

Job one: design what the agent does. Job two: build everything that makes it run - the servers, the security, the session management, the tool wiring, the monitoring. If those words mean nothing to you, skip to Section 2 and come back. If they mean everything to you, keep reading: Anthropic just eliminated that entire job.

That second job had nothing to do with your product. It took most teams three to six months. And you rebuilt it from scratch every time a new model shipped, because your harness encoded assumptions about what the old model couldn’t do.

Claude Routines are scheduled Claude Code cloud sessions. Prompt plus connectors plus trigger. No server. No agent loop. No infrastructure. Fires on a schedule or a webhook and runs until it’s done. Available on any paid Claude plan today.

Claude Managed Agents are a hosted production layer for shipping agents to users. You define the agent once - model, system prompt, tools, permissions. Anthropic runs the servers, sandboxing, session persistence, credential vault, tool orchestration, and observability.

Routines are for your own recurring work. Managed Agents are for things you ship to other people.

2. What this means even if you never write code

If you're not a developer, here's why you should care.

Routines connect to Gmail, Slack, Google Calendar, Notion, and the web. That means any recurring task you do with those tools can now run on autopilot: pulling competitor news every morning, prepping a summary before your Monday standup, flagging overdue items in your project tracker, or compiling weekly metrics from multiple sources into a single Slack message. You write what you want in plain English, pick a schedule, and it runs while you sleep.

Managed Agents matter because they change how fast your engineering team can ship AI features. If you're a PM writing a brief for an AI-powered feature, the timeline just collapsed from "3-6 months of infra work before we start" to "days."

The builds below show exactly what that looks like. Even if you skip the code prompts, read the use cases - they'll change how you scope your next roadmap.

3. Claude Routines

What a Routine actually is

A Routine is a saved Claude Code configuration: prompt, connectors, and a trigger. It runs on Anthropic’s cloud infrastructure as a full Claude Code session. Your laptop can be off.

Read this before you write any prompt in this section.

Every Routine run starts from zero. No memory of yesterday. If your Routine needs continuity (what changed since the last run, what issues are still open, what you already flagged), write today’s output to Notion at the end and read it back at the start of the next run. Every prompt below assumes you’re doing this.

Second thing that burns people: all connected integrations are available by default and Claude can use them without asking. Remove every connector the Routine doesn’t need before you schedule it. Less access, less surface area.

Three trigger types

1. Schedule — minimum one hour interval. Daily, weekly, or hourly. The standard case.

2. Webhook / API — HTTP POST to a per-Routine endpoint with a bearer token. Research preview. Good for event-driven flows.

3. GitHub — fires on PR open, push, release. Each matching event starts its own session. Good for code-adjacent automation.

The difference from a cron job running a script: the session reasons. It doesn’t execute a fixed procedure. It reads output, makes judgment calls, handles edge cases mid-run. A bash script can’t decide whether a dependency vulnerability is worth waking someone up for. A Routine can.

Routine 1: Competitor intelligence brief

If you manage a product, run marketing, or lead a team, start here.

Every morning before you open Slack, this Routine has already scanned your competitors’ websites, blogs, and social accounts for anything new. Product launches. Pricing changes. New positioning. Job postings that signal where they’re investing. It writes a short brief and drops it in a Slack channel or Notion page.

Connectors needed: Slack (or Notion). Web browsing is built in.

Go to claude.ai/code/routines. New Routine. Set trigger to Schedule, daily at 7 AM. Paste this:

You are a competitive intelligence analyst.

Every morning:

1. Visit these competitor websites and blogs: [LIST URLS]

2. Search the web for news about these companies from the last 24 hours: [LIST COMPANY NAMES]

3. For each competitor, note:
   - Any new product announcements or feature launches
   - Pricing or packaging changes
   - New job postings that signal strategic shifts (e.g., hiring 5 AI engineers = investing in AI)
   - Blog posts, case studies, or press mentions

4. Read today’s context from: [NOTION PAGE URL or “yesterday’s Slack message in #competitive-intel”]

5. Compare against yesterday’s findings. Only report what’s NEW.

6. Post to Slack channel [CHANNEL ID]:
   Competitive Brief — [Date]
   [Company Name]: [What changed and why it matters in 1-2 sentences]
   If nothing changed: “No new competitor activity in the last 24 hours.”

7. Write today’s findings to [NOTION PAGE URL] so tomorrow’s run has context.

Keep the brief under 200 words. No filler. Only report things a product or marketing lead would act on.

Do not write anything back to me. Execute these steps immediately.

One known limit: web browsing finds publicly available content only. It won’t catch changes behind a login wall or gated announcements.

Hit New Routine:

Paste the Prompt with your details:

And it’s on:

Do not write anything back to me. Execute these steps immediately.

Routine 2: Weekly team digest

If you run a team and spend Monday mornings piecing together what happened last week, this replaces that.

Every Monday at 7:30 AM, this Routine reads your team’s Slack channels, Calendar events from the past week, and a Notion project tracker. It writes a digest: what shipped, what’s blocked, what meetings are coming up, and what needs your attention.

Connectors needed: Slack + Google Calendar + Notion.

You are a team operations assistant.

Every Monday at 7:30 AM:

1. Read messages from the last 7 days in these Slack channels: [LIST CHANNEL IDS]
2. Pull all calendar events from the last 7 days for: [LIST TEAM MEMBER EMAILS]
3. Read the current state of this Notion project board: [NOTION PAGE URL]
4. Write a digest with these sections:
   SHIPPED LAST WEEK: items that moved to Done on the project board
   BLOCKED: items flagged as blocked, with who owns them
   KEY DECISIONS: any decisions made in Slack threads (look for messages with reactions or explicit “decision:” language)
   THIS WEEK’S CALENDAR: important meetings coming up in the next 5 days
   ATTENTION NEEDED: anything that looks stalled, overdue, or unassigned

5. Post the digest to Slack channel [CHANNEL ID].
6. Keep it under 300 words. Use short sentences.

Do not write anything back to me. Execute these steps immediately. 

It pings you in Slack:

Routine 3: PR review on every push

If you ship code daily, this one's for you.

You push code. Nobody reviews it until standup. By then the context is gone.

This Routine fires on every new PR via GitHub trigger. Reads the diff. Checks for obvious issues - missing error handling, hardcoded values, undocumented functions, test coverage gaps. Posts a structured review comment directly on the PR before any human has looked at it.

Not a replacement for code review. A first pass that catches the easy stuff before a senior engineer has to.

Connectors needed: GitHub.

A note on model choice: Use Sonnet 4.6 for this, not Opus. Opus is slower and roughly 5x the cost per token. For a monitoring task that runs on every PR, Sonnet handles the reasoning fine and keeps your bill reasonable.

Go to claude.ai/code/routines. New Routine. Set trigger to GitHub, connect your repo. Paste this:

You are a code reviewer for [REPO NAME].

When triggered by a new pull request:

1. Read the full diff.
2. For every function signature, return type, or exception path that changed in the diff, trace the downstream callers in this repo. Identify any caller where the contract changed but the caller was not updated in the same PR.
3. For every new external call (API, DB, filesystem), check whether the surrounding code handles the failure mode it can actually produce (timeout, auth failure, malformed response). Not "is there a try/except" — does the handler match the real failure.
4. For every changed branch of logic, check whether an existing test covers the new behavior. If not, name the test file that should have been updated.
5. Also flag: hardcoded secrets, N+1 queries, and unbounded loops.
6. Post a review comment on the PR:


- No issues: "Automated review: no contract breaks, failure modes handled, tests cover changed paths. Ready for human review."
- Issues: list them by severity (contract breaks first, then failure handling, then test gaps, then style).

Keep the tone factual. No praise. No padding.
Do not write anything back to me. Execute these steps immediately.

Click Run Now on a test PR first. Confirm the comment appears. Then enable the GitHub trigger.

Scale limit to know: Step 2 ("trace the downstream callers") works well on repos under ~5,000 files. On large monorepos, Claude may time out or miss callers outside the context window. For repos that size, scope the review to the changed files and their direct imports only.

Routine 4: Dependency audit, weekly

If you maintain a production codebase, this catches the things you forget to check.

Your package.json has 47 direct dependencies. You have no idea which ones are outdated, which have known CVEs, or which haven’t been touched in three years.

Every Monday at 7 AM, this Routine reads your requirements file, checks each dependency against current stable versions, flags anything with a known CVE or more than two major versions behind, and posts a ranked report to Slack. Severity one is a security issue. Severity two is significantly outdated. Severity three is worth noting.

Connectors needed: GitHub + Slack.

You are a dependency security analyst for [REPO NAME].

Every Monday at 7 AM:

1. Read the requirements.txt or package.json from the main branch
   of [REPO URL].

2. For each dependency, check:
   - Current version in the file vs latest stable release
   - Any known CVEs against the installed version
   - Last release date of the package

3. Classify each finding:
   SEVERITY 1: Known CVE against installed version
   SEVERITY 2: Two or more major versions behind latest stable
   SEVERITY 3: No release in 18+ months (maintenance risk)

4. Post to Slack channel [CHANNEL ID]:

   Dependency Audit — [Date]

   SEVERITY 1 (fix now):
   - [package] [installed version] — CVE: [ID] — [one line description]

   SEVERITY 2 (update this sprint):
   - [package] [installed] → [latest]

   SEVERITY 3 (note for backlog):
   - [package] — last release [date]

   If nothing flagged: "Dependency audit complete. No issues."

Do not write anything back to me. Execute these steps immediately.

4. Claude Managed Agents

What the architecture actually is

Managed Agents has three components. You need to understand all three before you write a system prompt or brief an engineer.

1. The Agent is the brain. Model, system prompt, tools, MCP servers, permission policies. You define this once. It gets an ID. Every session references that ID.

2. The Environment is the hands. A cloud container with Python, Node.js, and Go pre-installed. Network access rules you configure. Mounted files. Anthropic provisions it, scales it, and tears it down. You define what it can reach.

3. The Session is the log. Append-only event stream. Every tool call, every decision, every output, timestamped and traceable. Persists through network disconnections. Survives for hours.

The brain/hands separation is what makes this infrastructure-safe. Claude generates code inside the session. That code runs in the sandbox. The sandbox never handles raw credentials - API keys are stored in a vault outside it and proxied in via MCP. A compromised prompt can’t exfiltrate your tokens.

Built-in tools available to every agent out of the box: bash, file read/write, web search, web browse, code execution, and MCP server connections.

Pricing: $0.08 per session-hour (idle time is free), plus standard token costs ($3/$15 per million tokens for Sonnet 4.6 input/output). A practical example: a 10-minute Sonnet session that processes 50K input tokens and generates 5K output tokens costs roughly $0.16 in compute plus $0.01 in session time. Most runs cost pennies.

Managed Agents Build 1: Bug-to-PR agent

If your team spends hours going from “bug identified” to “fix in review,” this collapses that to minutes.

This is what Sentry actually shipped.

Their debugging agent Seer identifies the root cause of a flagged bug. Previously that’s where it stopped: a diagnosis with no fix. They paired it with a Claude Managed Agent that reads the diagnosis, writes the patch, and opens a PR. Flagged bug to reviewable fix, zero human steps.

The numbers from Sentry’s writeup: over 1 million root cause analyses processed per year, near-immediate reviews on more than 600k PRs per month, and the initial integration shipped in weeks by a single engineer.

Two things worth noting about how they got there. First, they didn’t build the sandboxing or session layer. Anthropic runs both. Second, they kept the scope tight: the agent only handles bugs Seer has already diagnosed. It doesn’t diagnose. It doesn’t refactor. One job, one handoff, one PR.

The session runs for however long it takes. Could be four minutes. Could be forty. Anthropic handles the runtime either way.

Go to platform.claude.com, Managed Agents, Quickstart. Paste this into the “Describe your agent” box:

You are an autonomous bug fix agent for [REPO NAME].

When triggered with a bug report and root cause analysis:

1. Read the full context: error message, stack trace, root cause
   summary, and affected file paths.

2. Check out the affected files from the main branch.

3. Write a minimal fix that addresses the root cause without
   changing unrelated logic. Do not refactor. Do not optimize.
   Fix the specific bug.

4. Write or update the test that would have caught this bug.
   The test must fail on the unfixed code and pass on your fix.

5. Open a PR against main:
   - Title: "Fix: [error type] in [file]"
   - Body: root cause summary, what changed, how to verify

6. Post to Slack channel [CHANNEL ID]:
   "Bug fix PR opened: [PR URL]. Affected: [file]. Review
   required before merge."

Never merge without human approval. If the fix requires
changes to more than 3 files, post to Slack and stop:
"Fix scope too broad for autonomous PR. Manual review needed."

After launch: the scope limit (”more than 3 files, stop”) is the most important line. Remove it and you’ll get PRs that touch half the codebase. Keep it for the first month.

Managed Agents Build 2: Codebase security scanner

If your security audits are quarterly and expensive, this runs weekly for pennies.

Most security audits are quarterly, expensive, and done by people who don’t know your codebase. This agent runs weekly, knows your whole repo, and costs $0.08/hour.

Every Sunday night, it reads every file in your repo, maps the attack surface, cross-references against known vulnerability patterns, and outputs a structured report ranked by severity and exploitability.

You are a security analyst for [REPO NAME].

When triggered:

1. Read every file in the repository. Focus on:
   - Authentication and session handling
   - Input validation and sanitization
   - Database queries (SQL injection surface)
   - File operations and path traversal risks
   - Dependency imports with known CVE history
   - Hardcoded secrets, tokens, or credentials
   - API endpoints with missing authorization checks

2. For each finding:
   - Location (file and line)
   - Vulnerability class (OWASP category if applicable)
   - Severity: Critical / High / Medium / Low
   - Exploitability: how a real attacker reaches this
   - Recommended fix in one paragraph

3. Create a Notion page titled "Security Scan — [Date]" with:
   - Executive summary (3 sentences)
   - Findings table sorted by severity
   - Full details per finding

4. Post to Slack channel [CHANNEL ID]:
   "Security scan complete. [X] critical, [Y] high, [Z] medium.
   Full report: [Notion URL]"

Flag false positives explicitly rather than omitting them.
A false positive with explanation is better than a missed finding.

Scale limit: “Read every file” works on repos up to a few thousand files. For larger codebases, scope the scan to specific directories (auth, API routes, data layer) and rotate coverage across weeks.

Managed Agents Build 3: API migration agent

If you're deprecating an API and have dozens of services still calling it, this agent maps every call and writes the migration diffs.

You’re deprecating a v1 API. You have 47 services calling it. Nobody knows which endpoints each service uses.

This agent takes the old API spec and the new API spec. Reads every file across every service. Identifies every call to the old endpoints. Writes the migration diff for each service. Opens draft PRs.

You are an API migration agent.

Context:
- Old API base URL: [V1_API_URL]
- New API base URL: [V2_API_URL]
- API change summary: [PASTE DIFF OF BREAKING CHANGES]

When triggered:

1. Read every file in the repository looking for calls to
   the old API base URL or deprecated endpoint paths.

2. For each file with old API calls:
   - List every deprecated call found
   - Map each to its v2 equivalent based on the API change summary
   - Note any calls with no direct v2 equivalent (flag for manual review)

3. For files with clear 1:1 migrations only:
   - Write the updated file with v2 calls
   - Open a draft PR: "Migration: [filename] v1 → v2 API"

4. For files requiring manual review:
   - Create a Notion page: "API Migration — Manual Review Required"
   - List each ambiguous call with context

5. Post to Slack channel [CHANNEL ID]:
   "API migration complete. [X] files auto-migrated (draft PRs
   open). [Y] files need manual review: [Notion URL]"

Never merge draft PRs automatically. Human review required on every PR.

These three builds are engineering workflows, but the same architecture works for any domain: data quality checks across warehouse tables, content moderation pipelines, customer onboarding sequences, or compliance document review. The pattern is the same - define the agent, scope its tools, let it run.

5. What’s live, what’s preview, and what breaks

Public beta, available today: Routines on all paid plans (Pro 5/day, Max 15/day, Team and Enterprise 25/day). Managed Agents on the API - requires an API key, separate from your Claude chat subscription - with the built-in tool set (bash, file ops, web search, web browse, code exec, MCP).

Research preview, rate-limited: Webhook triggers on Routines. Multi-agent patterns where agents spawn sub-agents. Cross-session memory. Self-evaluation against success criteria.

Known limits that break builds:

• Routines are fully stateless between runs. No workaround except external storage.

• Webhook triggers don’t retry on failure. You need to handle at-most-once delivery on your end.

• Managed Agent sessions are capped at several hours. Long-horizon jobs need to checkpoint and resume.

• Compliance posture is still maturing. If you’re in HIPAA, PCI, or FedRAMP scope, check Anthropic’s trust center at trust.anthropic.com and confirm BAA/DPA availability before you design around this.

What to ship first

If you’re not technical: set up the competitor intelligence Routine or the weekly team digest. Takes an afternoon. Runs forever. Pro plan (5 runs/day) covers both. If you’re running 3+ daily Routines, upgrade to Max.

If you ship code: the PR reviewer and dependency audit are the fastest wins. Three Routines are running in my setup right now.

One thing nobody tells you: when a Routine fails, it fails silently. No Slack message, no email. Check your Routine’s run history at claude.ai/code/routines after the first scheduled run. If you see a red status, read the log, fix the prompt, and hit Run Now before re-enabling the schedule.

For Managed Agents, pick the build that addresses your biggest current bottleneck. Security coverage gaps: run the scanner weekly. Slow bug resolution: wire the bug-to-PR agent. API migration backlog: the migration agent clears it in one session.

Ship one. Review the output for four weeks. Then ship the next.

Final Words

The teams pulling ahead in the next year won’t be the ones with the most agents.

They’ll be the ones wiring these into permanent workflows, reviewing outputs weekly, and adjusting prompts the moment something drifts.

One Routine this afternoon. One Managed Agent this month. Review both for four weeks before you ship the next.

Subscribe now

PS. The Product Growth edition covers Routines and Managed Agents from the PM side. If that’s useful for your team, it’s at news.aakashg.com.

Beyond AI, here’s what I found interesting this week:

1/ Tennis players live 9.7 years longer than sedentary people in the Copenhagen City Heart Study, which followed 8,577 Danes for 25 years. Badminton added 6.2 years, soccer 4.7, cycling 3.7, swimming 3.4. The authors flag this as observational, so association not causation, but the ranking is striking: the sports with the most social interaction added the most years. You literally cannot play tennis alone. The forced social component is doing a lot of the work.

2/ France quietly pulled its gold out of New York and moved it to Paris. The official reason was “higher quality bars.” The real reason: they made €12.8 billion in profit on the swap and exited the U.S. financial system while calling it maintenance. Germany and Italy have $245 billion still sitting in Manhattan. Politicians in both countries are already asking for it back.

3/ Leopold Aschenbrenner sold every share of Nvidia last quarter and bought fuel cells, Bitcoin miners, and power companies instead. His logic: you can order 100,000 GPUs in six months. You cannot add 500 megawatts to the grid in six months. Oracle just signed a 2.8 GW fuel cell deal with his largest holding. The stock jumped 15% after hours. He’s 24.

4/ The flat stripe on your towel is called a dobby border. It’s the only part engineered to not absorb water. It exists to stop the edges from unraveling in the wash, control shrinkage, and give hotels a flat surface to embroider their logos on. The part that feels decorative is doing all the structural work.

That’s all for today. See you next week,

Aakash

P.S. Want my AI tool stack? Join my bundle. Want my job search coaching? Apply to my cohort.

Share

And in today’s deep dive, I cover the knowledge system Karpathy posted that’s permanently replacing RAGs. It’s the first thing I’ve tried that makes your AI knowledge compound instead of reset.

Subscribe now

Ariso: Your AI Chief of Staff

I had 80+ loose files rotting in my Google Drive root. Asked Ariso’s AI to organize them.

Twenty minutes later: GTM, Product, Finance, Legal, Content - all sorted, no rules or templates needed. It just read the files and figured it out. One less Saturday wasted on admin.

Try it free at ariso.ai/aakash.

Try Ariso

There’s a billion AI news articles every week. Here’s what actually mattered.

News

Anthropic just made every agent orchestration startup obsolete

Claude Managed Agents launched this week in public beta. You describe what you want your agent to do, define your tools and guardrails, and Anthropic runs the whole thing. Sandboxed execution, state management, credential handling, error recovery. All of it, handled.

What used to take months of infrastructure work now takes days. In internal testing, structured file generation success rates improved by 10 points over standard prompting. Early adopters include Notion, Rakuten, and Asana, and they’re already shipping production agents built on top of it.

Here’s what people are missing. Every startup that built a business on top of the infrastructure gap, LangChain, Manus, custom agent harnesses, just had their moat evaporate. Anthropic isn’t just offering a faster path to deployment. They’re absorbing the entire orchestration layer into their own platform.

The pricing tells the story. $0.08 per agent runtime hour, on top of model usage. That’s not an enterprise premium. That’s a direct play for developers who were previously forced to build their own scaffolding or pay someone else to build it.

Why this matters for you. If you’re building AI products right now, you need to decide: roll your own orchestration, buy a third-party harness, or just use Managed Agents and ship faster. The third option just got a lot more credible.

The way I see it, this is Anthropic pulling the rug out from under the middleware layer of the AI stack. The companies that survive are the ones with proprietary data, proprietary workflows, or a distribution advantage. Infrastructure players just lost a round.

Other News that Mattered this Week

• Google released Gemma 4 , the first time the Gemma family ships with a truly open commercial license. The 26B MoE model activates only 3.8B parameters at a time, and the edge models run on Raspberry Pi. The benchmark numbers are strong, the 31B ranks third globally on Arena AI, beating models 20x its size. But the license change is what actually matters. No usage restrictions buried in fine print, no legal review friction. Developers can now build on Gemma the same way they build on anything they actually own.

• Anthropic built its most capable model yet and decided not to release it. Claude Mythos Preview found a 27-year-old OpenBSD vulnerability, broke out of its sandbox during testing, and emailed a researcher to confirm it had done so. Instead of a public launch, Anthropic formed Project Glasswing: a restricted coalition of 40+ companies including Apple, Google, Microsoft, and Amazon with $100M in model credits to use Mythos exclusively for defensive security work. Responsible release or competitive gatekeeping? Probably both.

• Meta finally released the first model from their team led by Alexandr Wang, Muse Spark. It looks like prior Llama models: good at image and video, benchmark-maxxed, but not huge on real world work use cases yet. Data analysis and counting in images are amongst its top skills.

• Claude’s M365 connector is now available on every plan, including free. Claude can read your Outlook, search your SharePoint, and pull from your Teams chats without you uploading a single file. Microsoft Copilot charges $30/user/month for essentially the same access. Anthropic just made that a much harder sell.

Resources

• AutoAgent is the open-source library every agent builder needs to know about right now. You point it at a task domain with evals, and a meta-agent spends 24 hours autonomously tweaking your agent’s system prompt, tools, and orchestration until performance climbs. No human tuning. It hit #1 on SpreadsheetBench (96.5%) and the top GPT-5 score on TerminalBench (55.1%), beating every hand-engineered entry on both leaderboards. I’ve been watching this closely because it changes the question from “how do I tune my agent?” to “why am I tuning my agent at all?”

The AI Second Brain That Actually Works

Karpathy posted a knowledge base system that replaces RAGs and achieves what that Knowledge management system is not been able to do for years. I’ve been running it for a week and a half. It’s the first thing I’ve tried that actually makes your AI knowledge compound instead of reset.

That’s today’s deep dive:

1. Why every second brain you’ve built eventually died

2. What Karpathy actually figured out (and why it spread so fast)

3. Four use cases for engineers and builders

4. How to set it up in 30 minutes

5. What it looks like after 30 days

Subscribe now

1. Why Your Second Brain Keeps Dying

You hit a hard problem. You solved it. Three months later the same problem shows up in a different form and you can’t remember what you figured out the first time.

The folder exists. The notes exist. The solution is gone. It lived in your head and left when you moved on.

That’s not a discipline problem. Maintaining a knowledge base is a specific job: cross-referencing sources, updating summaries, flagging when new information contradicts old conclusions. You’re already running six things. The moment you fall behind by a week, the guilt of catching up is enough to make you never open it again.

Every AI tool has the same flaw underneath. NotebookLM, ChatGPT uploads, Notion AI: they all retrieve at query time instead of building anything persistent. Ask a question, get an answer, session ends. Tomorrow: scratch. Your knowledge never compounds. The AI never gets smarter about your domain.

The best engineers and builders don’t just know more. They forget less. And until now, no tool has actually helped with that.

2. What Karpathy Actually Figured Out

Karpathy hit the same wall with ML papers. Every session started from scratch. Ask a question connecting three papers, get the answer. Come back the next day: the AI had forgotten everything it built.

He tried every tool designed to fix this. Same result. So he asked a different question: what if instead of waiting until query time to process your sources, the AI compiled them once at ingest and maintained that compilation as new sources arrived?

He built that system. 18.7 million views on X. 5,000+ GitHub stars.

Three pieces make it work.

1. Raw is your junk drawer. Articles, transcripts, docs, notes. Dump it in without organizing. That’s the AI’s job.

2. Wiki is what the AI builds from raw. Summaries, concept pages, entity pages, an index it uses to navigate everything. You never write the wiki. The AI writes and maintains it entirely, every session.

3. Schema is a single CLAUDE.md file that tells the AI what your knowledge base is about and how to organize it. You write this once. The AI follows it every session.

The tedious part of maintaining a knowledge base is the bookkeeping. Humans abandon wikis because the maintenance burden grows faster than the value. The LLM doesn’t get bored. That’s the shift.

3. Four Use Cases

Use Case 1 - Stakeholder Memory

The wiki ends up knowing your CTO better than most new employees do after their first month.

Every meeting note, Slack thread, and email thread goes into raw/. The wiki builds a page per stakeholder: what they care about, what they’ve pushed back on, what language has landed, what they’ve approved and why.

Before your next meeting with the CTO, run this:

Pull the stakeholder page for [Name].
What have they pushed back on before?
What framing has landed with them?
What should I lead with tomorrow?

You walk in knowing exactly what objections are coming, what framing will land, and what to address defensively. Most PMs figure this out live, in the meeting. Often too late.

The relationship context you were never going to write down consistently is now written down. Every meeting. Every decision. Every signal. All queryable.

Use Case 2 - Side project context that doesn’t evaporate

You work on your main job five days a week. Your side project gets weekends. By the time you open it Saturday morning you can’t remember what you were building toward. You spend the first hour just getting back up to speed.

Drop your planning notes, decisions, and TODOs into raw/ after every session. Write a brief note before you close the laptop about where you left off. The wiki maintains your running context.

Saturday morning prompt:

Where did I leave off on [project]? 
What was I trying to solve and what's the next step?

You start building in 5 minutes instead of 60.

Use Case 3 - Team Onboarding That Doesn’t Lose Knowledge When People Leave

A senior employees leaves your team. Years of context walk out with them. Why you made that architecture call 18 months ago. What users said about the feature you cut. What the CTO actually cares about versus what they say they care about.

If the wiki exists, the transition is painful but survivable. If it doesn’t, the next employees starts over.

Drop past docs and meeting notes into raw/. A new employee joins, they read the wiki, they get 80% of the context you’d otherwise spend weeks transferring. Karpathy named this use case directly in his gist: “Business/team: an internal wiki maintained by LLMs, fed by Slack threads, meeting transcripts, project documents, customer calls. The wiki stays current because the LLM does the maintenance that no one on the team wants to do.”

That’s your second brain working at team scale. The institutional memory that usually walks out the door when someone leaves, finally written down somewhere the next person can actually find it.

Use Case 4 - Solutions you solved once and forgot

You spent four hours debugging a gnarly issue last month. You found the fix. You moved on. Now it’s back, in a slightly different form, and you’re starting from zero.

Drop your debugging notes into raw/ when you solve something non-obvious. The wiki builds a page per problem type. Next time it surfaces, you query before you debug.

Have I hit anything like [error / behavior] before? 
What was the fix and what caused it?

The answer is usually there. The problem is you never had anywhere to put it.

4. How to Set It Up

Path A: Use My Skill

Use this skill I’ve created for you:

Grab the Skill

Path B: Do it Manually

Create the folder structure:

mkdir -p my-knowledge-base/{raw/assets,wiki,outputs}
cd my-knowledge-base
git init && git add . && git commit -m "setup"

Four folders. Raw is source material the AI reads but never modifies. Wiki is what the AI builds. Outputs is where answers to queries live. Git gives you version history for free.

Write a CLAUDE.md that tells Claude what this knowledge base is for, how to categorize sources, and what to do when new sources contradict old ones. The key conventions: every wiki page cites its source, cross-references use [[page-name]] links, and an append-only log tracks every ingest.

Start ingesting one source at a time:

Read the schema in CLAUDE.md. Process [FILENAME] from raw/. 
Read it fully, discuss key takeaways with me, then: create a 
summary page in wiki/, update wiki/index.md, update all relevant 
pages, flag any contradictions with existing wiki content, 
and append an entry to wiki/log.md.

Read the wiki summary after each ingest. Ask one question. Do that five times and you already have 15-30 interconnected pages.

One thing to resist: the instinct to dump everything at once. Batch ingesting produces worse results. You lose the ability to guide what the AI emphasizes. Karpathy ingests one source at a time. So do I.

5. What 30 Days Looks Like

By week three, something shifts. You ask a question and the wiki answers with something you forgot you knew. A note from two months ago surfaces in response to a problem you’re facing today. A contradiction between two things you read six weeks apart gets flagged automatically.

That’s the compounding effect. It doesn’t happen in one impressive moment. It accumulates quietly, source by source.

After 30 days you have 50-80 interconnected pages built from your own work. Live pages per project, per library, per problem type. A growing collection of answered questions that represent your best thinking on the things you keep coming back to.

The wiki made me a better researcher because knowing it would compound made me more deliberate about what I put in. The raw folder is the cleanest set of notes I’ve ever kept.

Here’s where I land: the best builders don’t just know more. They forget less. This is the first system I’ve used that actually delivers on that.

That’s it for today’s deep dive. I wrote a full version of this for PMs in Product Growth. It includes a complete CLAUDE.md starter template you can drop straight into your project, and a setup skill which helps you the complete second brain architecture in just 60 seconds.

Beyond AI, here’s what I found interesting this week:

1/ Peanuts in Coke is one of the most accidentally perfect food pairings ever. The acid releases glutamate from the peanuts, generating umami in real time. The salt makes the sweetness hit harder. Georgia farmers invented this in the 1920s because they needed one hand free while working. They nailed the optimal flavor loop a century before food science could explain why.

2/ Apple accidentally built the world’s largest hearing aid company. AirPods Pro 2 got FDA clearance as a clinical-grade hearing aid in 2024. Prescription devices cost $4,700. AirPods cost $249. Every hearing aid company spent decades engineering invisible devices to reduce stigma. Apple made earbuds a status symbol first, then added the medical function as a software update.

3/ Nestlé’s response to the KitKat truck heist is the best unplanned marketing campaign of the year. 413,793 bars stolen somewhere between Italy and Poland. Still missing. Their move: fake presidential motorcade through Canadian cities. Security guards at retail displays. A job posting requiring “experience guarding high-value assets.” The stolen cargo is worth $620K. The Dexerto post hit 4.7M views. The thieves wrote the creative brief for free.

4/ Your morning coffee on an empty stomach is a cortisol bomb. Cortisol peaks when you wake up. Breakfast is the signal that pulls it back down. Skip it, add caffeine, and you’ve stacked two stress drivers before 9 AM. A meta-analysis across 399,550 people linked skipping breakfast to higher odds of depression. Your body uses that first meal as a hormonal reset. Skip it and your stress system runs in the background all morning like an app you forgot to close.

That’s all for today. See you next week,

Aakash

P.S. Want my AI tool stack? Join my bundle. Want my job searching coaching? Apply to my cohort.

Share

And in today’s deep dive I cover the 120+ features Anthropic has shipped in the last 90 days.

I rank them into S, A, B, C, and D tiers. Then I give you the key details on the S tier.

Subscribe now

Ariso: Your AI Chief of Staff

I am using an under-the-radar AI tool that talks to me in Slack - and it’s not OpenClaw. It’s Ariso:

Ariso connects to my email, calendar, and docs, then handles meeting prep, follow-ups, and the rest. It remembers every conversation and gets smarter the more I use it.

Built by cybersecurity founders from Durham NC (where I used to live). I personally recommend it. Best of all? It’s free right now: ariso.ai/aakash

Try Ariso

There’s a billion AI news articles every week. Here’s what actually mattered.

The Week’s Top News: Anthropic Leaked Claude Code’s Entire Source Code

The “safety-first AI lab” just had two major leaks this week: first, that its new Mythos model is so good it poses cybersecurity risks; and, second, ironically, the source code for Claude Code.

A source map file shipped inside the Claude Code npm package by accident. Chaofan Shou spotted it first and posted the download link on X.

Within hours, 512,000 lines of TypeScript were mirrored across GitHub, forked 41,500+ times, and picked apart by thousands of developers. Everyone is fixating on the embarrassment. But the real story is what’s inside the code.

I went through it, and there’s a lot. 44 feature flags for capabilities that are fully built but haven’t shipped yet. A persistent background agent called KAIROS that keeps working while you’re idle. An anti-distillation system that injects fake tool definitions to poison competitors trying to scrape Claude Code’s API traffic. And an “undercover mode” that strips all Anthropic references from commits when employees contribute to public repos.

Anthropic is preparing for an IPO, selling enterprise security, and positioning as the responsible AI company. Two leaks in one week make that pitch a lot harder.

The Other News That Mattered

News

• Sam Altman’s non-binding DRAM deals may have caused the worst consumer hardware crisis in a decade. He locked up 40% of global memory supply with letters of intent that were never real. DDR5 prices jumped 171%. Micron killed its 29-year-old Crucial brand to chase AI demand. MU is now down 33% from its post-earnings high. The signature was worth exactly what the paper said: nothing binding.

• Meta quietly released TRIBE v2, a model that predicts how your brain responds to anything you see, hear, or read. 70x higher resolution than v1. They open-sourced it, which tells you the research advantage is already captured. A company spending $135B in capex didn’t build this for academic citations.

Resources

• Jack Dorsey published From Hierarchy to Intelligence. The thesis: corporate hierarchy exists to route information, and AI can now do that continuously at scale. Block cut 4,000 jobs and framed it as acceleration, not cost reduction. Fascinating read. Whether you agree with Dorsey or not, every leader should read this and decide what it means for their org.

New Tools

• Google AI Quests is expanding to eight new languages. It’s a free, gamified AI literacy experience built with Stanford that teaches students aged 11-14 how AI actually works through real research scenarios like flood forecasting and disease detection. If you have kids or work in edtech, worth bookmarking.

Fundraising

• OpenAI closed $122 billion at an $852B valuation, the largest private funding round in history. But look at who wrote the checks: Amazon ($50B), Nvidia ($30B), SoftBank ($30B). All three are also OpenAI’s biggest vendors. The capital isn’t just flowing in, it’s circulating. $35B of Amazon’s piece is contingent on an IPO or AGI by 2028.

• WHOOP raised $575M at a $10.1B valuation. A screenless rubber band with SaaS metrics: 2.5M members, $1.1B bookings run rate, 103% YoY growth, cash flow positive. Qatar Investment Authority, Mubadala, Abbott, and Mayo Clinic all wrote checks. Sovereign wealth funds and healthcare institutions don’t invest $575M because recovery scores are cool. They’re buying a distribution channel for continuous patient monitoring outside the clinic.

I Ranked Every Feature Anthropic Shipped This Quarter

Anthropic shipped 120+ features in 90 days.

Most people saw a few headlines and moved on. Agent Teams, Opus 4.6, maybe Dispatch. But they’re missing the bigger picture. These features aren’t isolated launches. They build on each other in ways that aren’t obvious if you’re only watching one product line.

I researched every Anthropic release so far this year. Blog posts, changelogs, tweets from the engineers who built each feature, official docs, community reports. And I tested them.

Honestly, most of it is incremental. But about 30 features are worth your time.

Here’s what you need to know:

The Full Picture

I mapped every release to a calendar. Here’s what Q1 looked like for Claude:

That’s 40+ Claude Code releases, 15+ Cowork updates, 20+ API changes, and two new models in 90 days.

The thing that stands out when you see it on a calendar is the overlap. Agent Teams and Opus 4.6 shipped the same week. Code Review landed on a Monday, and by Friday they’d added 1M context GA and four more Claude Code features. No product line was waiting for another.

That matters because these features build on each other. /loop is more useful when the model can stay on task for 14 hours. Dispatch gets more powerful when it can launch Claude Code sessions. Worktrees work better when 1M context means each agent can hold the full repo. The product lines connect in ways that aren’t obvious if you’re only watching one.

I went through all of it. Here’s the tier list:

Now let’s cover everything in S tier.

S Tier: Claude Code

Auto Mode

Auto Mode lets Claude make permission decisions on your behalf. It runs two safety checks before every action.

I enabled it and gave Claude a refactor task: dependency injection across a user service, update all tests, make sure CI passes. Claude edited 14 files, ran tests 3 times, fixed two test failures, and committed. Zero permission prompts. 8 minutes total. Before Auto Mode, I would have clicked “Allow” roughly 40 times during the same task.

claude --enable-auto-mode

Not the same as --dangerously-skip-permissions. Auto Mode still won’t delete your database. It just stops asking permission to edit a CSS file.

Opus 4.6

Opus 4.6 launched in February and the world hasn’t remained the same ever since.

Everyone focused on the benchmarks. The part that matters is the task horizon. Opus 4.6 can stay on a single task for over 14 hours without losing track. That’s what makes /loop, Agent Teams, and Dispatch viable on real projects.

Without this model, half the features on this list break on anything complex.

Agent Teams

Agent Teams lets a lead agent delegate to up to 10 parallel teammates that coordinate through a shared task list.

I tested this on a side project with three open issues: an auth flow, a dashboard layout, and a broken CSV export. Described all three to the lead agent and let it delegate. All three had PRs open within 40 minutes. Two passed tests immediately. The third needed one manual fix on a type mismatch.

Add CLAUDE_CODE_EXPERIMENTAL_AGENT_TEAMS=1 to your environment and describe multiple tasks in a single prompt. Requires Opus 4.6. Research preview.

Cowork Launch

Cowork is Claude running persistently on your computer with access to your apps, files, and screen. You message it, walk away, and come back to finished work. It’s a great skin on Claude Code:

Before a 2-hour meeting, I told Cowork: “Catch up on my Slack DMs and summarize anything I need to respond to. Then find the Q1 metrics deck in my Drive and update the revenue chart with this month’s numbers.”

When I got back, Cowork had summarized 12 Slack DMs into three priority buckets, flagged two that needed immediate replies, and updated the chart. The Slack summary saved me 30 minutes of scrolling.

Connectors

Connectors let you plug Claude into your work tools with one click. Gmail, Slack, Notion, Figma, Asana, Google Drive, and 50+ others as of February 2026. Built on MCP, but you don’t need to know that. You just toggle it on.

I connected the Figma connector and pointed Claude Code at a mockup for a settings page. It read the frame, extracted the layout and tokens, and generated a React component that matched the design. When I asked it to adjust spacing, it updated both the code and pushed the changes back to the Figma file. Design-to-code and code-to-design in one loop.

Connectors work across Claude.ai, Cowork, and mobile. Any MCP server anyone builds becomes a connector you can add. Refer docs here.

/loop

/loop runs a task on a repeating interval inside your session. These are the patterns I’ve been using:

/loop 5m /babysit
/loop 30m /slack-feedback
/loop 1h /pr-pruner

The first auto-addresses code review comments and pushes PRs to production every 5 minutes. The second puts up PRs for Slack feedback every 30 minutes. The third closes stale PRs hourly. These patterns come from the Claude Code team’s own workflow.

Loops run for up to 7 days and support up to 50 tasks per session. /loop can invoke other slash commands, so you can chain it to a skill and the whole workflow runs without you touching the terminal. I covered the origin of this pattern in Ship Features While You Sleep.

Final Words

Start with three things today. Turn on Auto Mode so you stop clicking “Allow” forty times per session. Set up Cowork so Claude works while you’re in meetings. And run /loop 5m /babysit on your next PR.

I’m shipping faster and I’m not working more hours. Not every feature here will matter for your workflow, but the S tier ones changed mine.

That’s it for today’s deep dive.

I wrote an extended version for paid Product Growth Subscribers covering A, B, C, and D Tier features. Read it on Product Growth →

Beyond AI, here’s what I found interesting this week:

1/ Zuckerberg’s yacht once sailed 9,600 nautical miles to the South Pacific and waited two months for him to show up, but he never boarded and the vessel turned around empty. It burns 1,165 gallons of diesel per hour and in nine months produced what 400 American households emit in an entire year. Meta pledges net zero across their value chain by 2030. Net zero is a line item in someone else’s budget.

2/ The S&P 500 had its worst quarter since 2022. Forward P/E fell to 19.7x, below the 5-year average. But every time it’s dropped below 20x since 2020, forward returns have been positive. Miss just the 10 best trading days and your returns get cut in half. The worst time to look at your portfolio is the best time to add to it.

3/ GTA 6 isn’t delayed because $3B is hard to spend. GTA 5 still generates $700 million a year thirteen years after launch, and day one of GTA 6 is also the last day of GTA Online’s current economy. Every delay is Rockstar protecting the most profitable entertainment product ever made while making sure its successor can match that run rate on arrival.

That’s all for today. See you next week,

Aakash

P.S. Want my AI tool stack? Join my bundle. Want my job searching coaching? Apply to my cohort.

Share

Everyone is comparing it to OpenClaw and Claude Cowork. But Computer is doing something different.

I’ve been testing it for weeks. Today’s deep dive is your complete guide to it.

Ariso: Your AI Chief of Staff

I had 80+ loose files rotting in my Google Drive root. Asked Ariso’s AI to organize them.

Twenty minutes later: GTM, Product, Finance, Legal, Content - all sorted, no rules or templates needed. It just read the files and figured it out. One less Saturday wasted on admin.

Try it free at ariso.ai/aakash.

Try Ariso

There’s a billion AI news articles every week. Here’s what actually mattered.

The Week’s Top News: OpenAI has left the video generation business

OpenAI just exited the video generation business entirely. App dead. API dead. No video inside ChatGPT. Disney’s $1 billion deal, signed four months ago, is dead.

Read that again. This isn’t a consolidation into the super app. Altman told staff Tuesday that OpenAI is winding down all products using video models. Disney’s own statement says they respect OpenAI’s decision to “exit the video generation business.” The Sora research team is being redirected to robotics.

The reason is sitting right there in the competitive data. Anthropic hit $19 billion in annualized revenue by early 2026 selling text and code. No video generation. No image generation. No consumer social app. No Disney deal. One product surface: chat, code, computer use, all in one place. OpenAI looked at where every dollar of market growth was coming from and saw the answer: coding and enterprise.

So now they’re copying the model. ChatGPT, Codex, and the browser merge into one app. Instant Checkout killed today too. Every consumer experiment is getting cut. What remains is the Anthropic playbook: one app, code and chat, enterprise and developer focus.

The Sora numbers explain the urgency. Total consumer revenue across iOS and Android since September: $1.4 million. Peak month was $540,000. Every video generation burned GPU compute that could have been running inference for ChatGPT or Codex instead. OpenAI’s own head of Sora announced generation limits because chips couldn’t keep up. At $14 billion in projected 2026 losses, every GPU matters.

Google just inherited the AI video market by default. Nano Banana already lives inside Gemini. No standalone app to manage, no separate brand to support. Among the majors, they’re the only ones left. Runway, Kling, Minimax, Luma, and the other independents are still shipping, but none of them have Google’s distribution.

Disney put $1 billion in stock warrants on a product that lasted six months. The deal was announced in December. Characters from Marvel, Pixar, and Star Wars were supposed to be generating fan videos on Sora by now. Instead, Disney is writing a polite press statement about “respecting OpenAI’s decision” while its legal team unwinds a deal that never produced a single licensed video.

Four months from billion-dollar partnership to obituary. That’s how fast the AI product landscape reprices when the unit economics don’t work.

The Other News That Mattered

• Anthropic shipped four features in 30 days that reverse-engineer every OpenClaw capability into an enterprise-safe product. Dispatch, Claude Code Channels, computer use in Cowork, curated plugin marketplace. The agent race rewards shipping velocity over hiring velocity.

• Someone poisoned LiteLLM, the Python package that routes AI API keys for NASA, Netflix, and Stripe. 97 million downloads a month. A single pip install exfiltrated every SSH key, cloud token, and .env file on the machine. The only reason it got caught is the attacker vibe coded the malware so badly it crashed computers before it finished. If you’re shipping AI agents, your supply chain is your attack surface.

• Apple put AI coding agents in Xcode 26.3, then blocked Replit and Vibecode from updating in the App Store. Replit dropped from first to third in developer tools downloads in two months. Every app built through vibe coding that ships as a web app is revenue Apple never touches. The policy isn’t about code execution rules from 2009. It’s about who gets to be the on-ramp.

• Google Research published TurboQuant, a compression algorithm that reduces LLM key-value cache memory by at least 6x and delivers up to 8x speedup on H100 GPUs. The internet is already calling it Pied Piper. Cloudflare’s CEO called it Google’s DeepSeek moment. The paper will be presented at ICLR 2026 next month.

Resources

• If you’re using Claude Code without plan files, you’re doing it wrong. Matt Van Horn’s piece on every Claude Code hack he knows is the best single resource I’ve seen this month. I have 70 plan files and 263 commits in the last 30 days. Plan first, build second.

Tools

• Claude Code cloud scheduling is the feature I've been waiting for. Set a repo, a prompt, a cadence. Your laptop can be off. Nightly dependency audits, stale PR reviews, test reruns, all running at 3am on Anthropic's infra. $200/month.

• MCP Apps hit mobile this week. Figma, Canva, Amplitude, all inside Claude on your phone. You stop opening apps. You open Claude and Claude opens them for you. The race to own the AI routing layer on your phone is now a three-way fight.

• Dispatch lets you run multiple Cowork sessions from your phone in Claude Code. Spawn a research task, a draft, and a gap analysis from the same thread while you're walking your dog. Each task runs independently on your desktop. Your phone is the command chair.

I'm doing a free masterclass on April 7th covering the two interview rounds that eliminate most PM candidates, behavioral and AI PM, and exactly how to pass them. Register here.

Register here

The Complete Guide to Perplexity Computer

The agentic AI world is moving fast. First it was Manus, Openclaw, Claude Code, then Claude Cowork. Now Perplexity has dropped Perplexity Computer.

They all want to be the tool you can’t live without. But how do you know which one to pick for your work? And how do they actually compare when you put them to work?

I have a feeling Computer is about to earn the same loyalty.

That’s what this guide is about. Perplexity Computer treats vague instructions as permission to do far more than you asked. And that costs real money. I spent my first week learning that the hard way.

But once I figured out how the system actually works, I built a prompting structure that controls the cost before Computer starts spending. And I built workflows that genuinely replaced hours of manual work.

I’m not here to inflate AI hype. I’m here to share what actually works.

Here’s what you need to know:

1. What Is Perplexity Computer

2. How to Use Perplexity Computer

3. Honest Limitations and Structural Risk

Subscribe now

1. What Is Perplexity Computer

Perplexity Computer launched on February 25, 2026. CEO Aravind Srinivas calls it a “general-purpose digital worker.”

Despite its name, it’s not a physical computer. It’s a cloud-based system made up of multiple AI agents working together.

It can research, design, build, test, deploy, and automate. With 400+ app integrations, it can also read your Gmail, push code to GitHub, query your Snowflake database, post to WordPress, and about 396 other things I’m still discovering.

All from a single prompt and a single interface. No terminal. No API keys. No local setup.

That last part matters. If you’ve tried OpenClaw , you know the headache. Local installations, environment configs, dependency hell. Computer runs entirely in the cloud. You describe the outcome. Computer handles everything else.

The 19 Models Under the Hood

I was jumping between Claude, Gemini, and GPT constantly. Different tabs, different contexts, different formatting. Copy-pasting outputs from one model into another just to get a complete workflow done. It was exhausting.

Computer eliminates that entirely. It sits on top of 19+ frontier AI models. Claude Opus 4.6 handles the core reasoning. Gemini runs deep research tasks. ChatGPT 5.3 assists with long-context recall. Grok takes the lightweight, quick tasks. Nano Banana generates images. Veo 3.1 produces videos (Google’s models).

Computer picks the best combo for the job. You write one prompt. Computer routes the work.

Tip: You can manually assign which models handle which parts of your task. Most people skip this and let Computer pick the most expensive option. Learning to control model routing is the fastest way to cut costs.

The Business Model Shift

Perplexity spent three years building a search engine at $20/month flat. Computer layers a consumption-based product on top of that entire base. Max subscribers get 10,000 credits per month. Usage-based pricing. Spending caps. You choose which models run your sub-agents.

This is a fundamentally different business than search. Search is a flat subscription. Computer is metered compute. That changes how you use it, how you budget for it, and how you think about what’s worth automating.

How Tasks Actually Run

Everything in Computer is organized around Tasks. A Task is not a conversation. It’s a job.

When you submit a prompt, the orchestrator (powered by Claude Opus 4.6) reads it, breaks the objective into subtasks, creates a separate sub-agent for each one, assigns each to the best-suited AI model, runs them in parallel, and compiles results in the right panel.

Every sub-agent consumes credits. More sub-agents means higher cost. This is why vague prompts are expensive. The orchestrator creates sub-agents for every plausible interpretation. Specific prompts create fewer, more targeted agents.

You can see the decomposition happening in the output panel as it runs.

Tasks run asynchronously in the cloud. Close your browser. Go to dinner. When you come back, everything is done.

Scheduled tasks take this further. Your machine doesn’t need to be on. Set it up once. Computer runs it on a recurring schedule whether your laptop is open, closed, or sitting in a drawer. This kills the need for OpenClaw to be honest, as Computer has zero local setup.

The Sandbox

Every task runs inside an isolated Firecracker microVM (same tech AWS built for Lambda). Real filesystem, real browser, real tool integrations. Complete isolation between tasks.

Computer can’t touch your local files. Everything runs in the cloud. If a task goes wrong, the failure stays contained. That’s the safety advantage over OpenClaw, which runs locally with direct access to your filesystem.

The sandbox also has advanced anti-blocking architecture. It navigates around anti-bot blockers that normally stop AI agents from browsing the web. That’s why it can scrape competitor sites and interact with web apps other agents can’t.

Limitations: no desktop app control, no screen/mouse access, file access limited to uploads and connectors.

Setting Up

You need a Perplexity Pro subscription ($20/month), Max ($200/month), or Enterprise Max ($325/seat/month). Pro gets 4,000 credits per month. Max gets 10,000. At launch, Max users received a one-time bonus of 20,000-35,000 extra credits (valid for 30 days).

Getting started takes 30 seconds. Go to perplexity.ai, log in, click the Computer icon in the left sidebar. You’re in. No local setup. No terminal. No environment variables.

Credit counter in the sidebar updates in real time. Usage button in the top right of any task shows its specific cost. Full breakdown at perplexity.ai/account/usage.

Connectors (And Why They’re the Real Product)

This is where Computer stops being a research tool and becomes a workflow execution system.

Connectors let Computer access your actual data and take real actions in your existing software. Not summaries. Real read-and-write access.

Setting them up takes about a minute per app. Click Connectors in the sidebar, browse the list, click Enable, complete OAuth. Done. Computer can now read from and write to that service.

Here’s what makes Computer’s connectors different from other AI tools. Most agents give you 10-20 integrations focused on productivity basics. Computer has 400+, and the interesting ones go way beyond Slack and Google Drive.

What you can actually do with these connectors.

Here’s what surprised me about these connectors. They’re not just productivity integrations. Some give you access to things that normally cost thousands or require engineering to set up.

• I asked Computer to build me an investment memo on Shopify. It pulled live financials from FactSet, earnings transcripts from Quartr, analyst sentiment from finance Twitter. No API keys. No data subscriptions. Just built-in access to FactSet, S&P Global, Coinbase, CBOE, Nasdaq, NYSE. All native.

• Through Plaid, you can link your actual brokerage and Computer builds you a live portfolio dashboard. Daily P&L, position cards, news feed, earnings calendar, price alerts. A personal Bloomberg terminal on a private URL. Full setup and prompt in Use Case 4.

• Research access is wild. CB Insights, PitchBook, Statista. Computer bypasses the paywalls. A $5,000 market sizing report? Computer just pulls it.

• The Snowflake connector made me stop and think. Connect it, and Perplexity auto-generates a Data Map of your schemas. Type “What were the top 10 customers by revenue last quarter?” Computer writes the SQL, runs it, hands you charts. Plain English in, data out. Same with Salesforce, HubSpot, Databricks.

• GitHub lets Computer write, refactor, and push code to your repos. Gmail, Slack, LinkedIn go deeper than you’d expect. I had it scan my inbox, summarize my newsletters, draft unsubscribe requests, and calculate my SaaS burn rate from receipt emails. Tag Computer in a Slack channel and it runs research without you switching tabs.

• Polymarket was unexpected. Live prediction market odds for industry events. Real-time sentiment without refreshing a dozen tabs. Reddit for early business signals. Toggl Track for time tracking. And this caught my eye: direct purchases through PayPal. An AI agent that can buy things. That’s new.

• For SEO, Ahrefs, Google Search Console, Google Analytics, WordPress, Cloudflare. Full audit, competitor analysis, keyword mapping, content drafting, and live site updates. One prompt.

• And the one that surprised me most. Computer connects to Amazon’s Selling Partner API and Shopify natively. If you’re running D2C, that changes everything.

Bring your own tools. If Perplexity doesn’t have a connector for something you use, provide an MCP (Model Context Protocol) server URL. This lets you hook Computer up to proprietary CRMs, custom analytics servers, or private APIs. Enterprise admins can share custom connectors across the organization.

I wrote about how Computer replaced $225K/year in marketing tools. There are 15,384 MarTech tools on the market right now, and the average marketing team uses only 33% of their stack’s capabilities. Companies are paying more for tools they use less every year. That’s the gap Computer is attacking. The argument is that the entire category is waste.

Skills and Custom Instructions

A Skill is a saved set of instructions that auto-activates when Computer recognizes a matching task. Same concept as Claude Skills or Cowork plugins.

Without Skills, you’re re-explaining your brand guidelines, formatting preferences, and reporting structure on every task. With Skills, you explain it once.

Computer ships with built-in Skills for Slides, Research, Research Report, and Chart. To create a custom one, click Skills in the sidebar, click + Create skill, and upload a .md file.

Skills stack. A research Skill can hand off to a report-formatting Skill, which hands off to a slides Skill. One prompt, full pipeline.

Tip: If you’ve explained the same thing to Computer twice, it should be a Skill. Brand guidelines. Formatting rules. Research methodology. Build it once, never repeat yourself.

Custom Instructions are different from Skills. Skills activate for specific task types. Custom Instructions apply to ALL tasks, all the time.

Keep them under 1,500 characters. The single most effective one I’ve found:

“Always come back to me and clarify any misunderstandings and challenge my thinking to make sure you’re very clear on the stated outcome, and let’s create a brief short plan before we build anything.”

That forces Computer to verify what you want before spending credits. One line. Massive savings.

The Perplexity Stack: Use the Right Tier

This is the part most people miss. Perplexity has four product tiers. Using Computer for a question Pro Search handles in 10 seconds is like hiring a full project team to send one email.

Perplexity Comet is the AI-native browser. Unlimited on Comet Pro. Tab-level browsing, page summaries, routine web research. 80% of daily AI work should happen here.

Pro Search answers factual questions quickly from a limited source pool. If the answer exists and you just need it collected, start here.

Deep Research synthesizes insights across 50+ sources into long-form reports. 20 runs per month on Pro. One run equals one serious research task.

Perplexity Computer is for multi-step projects that require parallel agents, tool integrations, code execution, and structured deliverables.

2. How to Use Perplexity Computer

You do not need to know how to code, manage API keys, or use a terminal. You write your stated intention in plain English. Computer autonomously researches the data, writes the code, and deploys a fully interactive web application with a live, shareable link.

Use Case 1: The Lead Machine

Instead of manually sourcing leads, Computer acts as a full-time SDR.

The workflow: For cold outbound, give it a list of 30 target companies. It researches each one, identifies the right decision-maker (bypassing the CEO to find the actual department head), and drafts 30 personalized emails referencing specific company milestones. It schedules follow-up sequences, leaving everything in your Gmail drafts for one final review.
Cold outbound prompt:

Here's a list of 5 companies I want to sell to: [LIST]. For each 
one, find the founder or relevant department head on LinkedIn (not 
just the CEO, find the actual decision-maker for [YOUR CATEGORY]). 
Research their company's recent news, funding, and pain points. 
Draft a hyper-personalized cold email for each one referencing 
something specific about them. Each email must reference at least 
one specific company detail. Schedule 3-day and 7-day follow-up 
sequences. Leave everything in my Gmail drafts for final review. 
Do not send any emails without my approval. Show me the company 
profiles and identified contacts before you start drafting.

PS: If any of the connectors are missing, computer will ask you to connect on the fly itself, making the configuration super easy.

View Full Task On Computer

Use Case 2: “Senior Advisor” Contract Auditing

Legal review means expensive billable hours. Computer catches blind spots before the lawyers even see the document.

Upload a partnership agreement or proposal. Computer reviews it like a senior consultant. It fact-checks claims against live web data. If a proposal claims a market grew by 43%, Computer checks whether it actually grew by 4%. It flags risky language and delivers a marked-up Word document with track changes.

Review this uploaded partnership proposal. Fact-check everything 
against live web data, identify deal terms, flag any vague or risky 
language, suggest improvements, and give me a top 5 concerns 
summary. Focus on financial terms, liability clauses, termination 
conditions, and any factual claims about market size or growth 
rates. No images, video, or code. Deliver it all in a Word document 
with track changes and comments. Before you start, propose your 
review approach and wait for confirmation. Flag any claims you 
cannot verify immediately.

Use Case 3: The Silent 8 AM Competitor Ping

Set up a recurring workflow that checks five competitors’ websites and X accounts every morning at 8:00 AM. It watches for pricing changes, feature drops, stealth launches, and brand mentions.

If nothing changed, it stays completely silent. No “all clear” messages. It only pings you when something strategic actually shifted.

Every morning at 8:00 AM EST, check these five competitors' 
websites for pricing changes, new feature announcements, and any 
new blog posts: [COMP 1], [COMP 2], [COMP 3], [COMP 4], [COMP 5]. 
Also search X for any mentions of them. If anything has changed 
since yesterday, send me a push notification with a summary. If 
nothing has changed, stay quiet. No daily summaries. No "all clear" 
messages. Maximum 5 items per report. Under 200 words. Only notify 
me if at least 2 qualifying changes are found.

View Full task on Computer

Use Case 4: Your Personal Bloomberg Terminal

This is one of the most impressive things I’ve seen Computer do. And you don’t need to write a single line of code.

The idea is simple. You connect your actual brokerage account, and Computer builds you a personalized finance terminal that runs on a private URL, updated automatically on whatever schedule you set.

Here’s the full flow.

Step 1: Connect your brokerage. Link Robinhood, Schwab, Fidelity, or whatever you use through Computer’s Plaid connector. Just say “connect my brokerage” and Computer walks you through OAuth. It never sees your login credentials directly. Once connected, it reads your holdings, positions, and transaction history.

Step 2: Build the terminal. Tell it “Build me a live portfolio dashboard that tracks my holdings.” Computer pulls your real portfolio data from the brokerage, fetches live market data using its native finance tools (FactSet, SEC filings, Coinbase, Quartr, all without API keys), and builds an interactive web dashboard deployed as a website.

Step 3: Make it always-on. The dashboard gets deployed with a private URL you can bookmark. Computer sets up recurring tasks to refresh data hourly, check for price alerts, earnings surprises, or significant moves. It sends you notifications when something noteworthy happens. A stock drops 5%. An earnings report lands. A position crosses your threshold.

Three things you need to do.

1. Connect a brokerage. Say “connect my brokerage.”

2. Describe what you want. “Build me a portfolio dashboard with my holdings, daily P&L, and news for each stock.”

3. Set up monitoring. “Check my portfolio every hour and notify me if any position moves more than 3%.”

No code. Computer handles data fetching, site building, deployment, and scheduling.

The build prompt:

Build me a live portfolio dashboard that tracks my connected 
brokerage holdings. Include current value, daily P&L, allocation 
breakdown, individual position cards with cost basis and gain/loss, 
a news feed for stocks I hold, an earnings calendar for my 
holdings, and configurable price alerts. Deploy as a private website 
I can bookmark. Do not make it publicly accessible. Dashboard only, 
no trading recommendations or financial advice. Before you build 
anything, propose the layout and wait for my confirmation. Show me 
a working preview before final deployment. Do not make it public 
without my approval.

Then follow up to make it always-on then give command to the task:

Set up a recurring task that refreshes my portfolio dashboard data 
every hour. Send me a push notification if any single position 
moves more than 3% in either direction, or if an earnings report 
drops for any stock I hold. If nothing noteworthy happened, stay 
quiet. Include the position name, change amount, and one line of 
context in each notification.

Use Case 5: Single-Ticker Investment Memos

If you’re exploring an acquisition, evaluating a stock, or preparing for a board conversation, this workflow replaces hours of manual research.

You can prep for a conversation about a potential investment. Would usually take an entire afternoon pulling data from 6 different sources. Computer hands a finished memo with charts in about 15 minutes.

View Complete Task on Computer

Build me an investment research memo on [TICKER SYMBOL]. Pull their 
latest financials and earnings transcript highlights. Compare their 
margins and growth to [COMPETITOR 1] and [COMPETITOR 2]. Check what 
analysts and finance Twitter are saying. Compile into a polished 
PDF with charts. I want a bull case, a bear case, and your final 
assessment. Maximum 10 financial sources. No video or audio. 3,000 
words max. After gathering the financials, show me the data summary 
before you start writing. If the same error happens twice, stop 
and ask me.

Use Case 6: Self-Auditing Your Marketing and SEO

Computer can turn the audit lens inward on your own assets.

Build a Skill defining great marketing for your brand. Computer crawls your website and scores every page. Instant visibility into what needs rewrites.

For SEO, give it your business details and let it run a full audit: Google Search Console data, competitor analysis, keyword mapping, content drafting, and technical fixes. All from one prompt.

Run a comprehensive SEO audit of [YOUR WEBSITE URL]. Compare to 
the top 3 competitors for [YOUR PRIMARY KEYWORDS]. Identify content 
gaps and draft a prioritized action plan. Use my connected Google 
Search Console and Ahrefs data. Focus on technical issues, content 
gaps, and keyword opportunities. Give me the action plan in order: 
technical fixes first, then content opportunities, then keyword 
targets. 3,000 words max. Do not make any changes to the website 
without my explicit approval. Before you start, propose the audit 
scope and wait for confirmation. Show me the technical findings 
before moving to content analysis.

Use Case 7: Podcast to TikTok Clip

Turning a long podcast into a short social clip normally means opening the video, downloading, finding the segment, trimming, converting to vertical, captioning, and exporting. Multiple tools. Multiple hours.

One prompt. Computer downloads the podcast, finds the segment by topic, extracts the clip, converts to vertical, captions it, and renders the final output.

Download the latest [PODCAST NAME] episode from [PODCAST URL]. 
Find the segment where they discuss [TOPIC]. Extract a 60-90 
second clip, convert it into vertical format for TikTok, and add 
captions with white text on a black background bar. Do not upload 
or publish anywhere without my approval. Show me the identified 
segment timestamp before extracting. Show me a preview before 
the final render.

View the Task

Use Case 8: Build Internal Tools in Plain English

Need a client intake form, a project tracker, or a simple CRM? Describe it. Computer builds it, styles it, and gives you a live link.

For tech users: Computer uses GPT-5.3-Codex to build full-stack apps with React frontends and Python backends, and pushes code directly to GitHub.

Build me a client intake form that collects a name, email, project 
type (dropdown with [TYPE 1], [TYPE 2], [TYPE 3]), budget range, 
and timeline. Save everything to a spreadsheet and send me a Slack 
notification every time someone submits. Provide a live shareable 
link. Single HTML file with inline CSS and JS. Clean, minimal 
design. Mobile responsive. Propose your approach before coding and 
wait for my confirmation. Show me the working form before you 
connect it to Sheets or Slack. Do not deploy without my approval.

How to Prompt Without Wasting Credits

I learned this the expensive way. Here’s the system I built so you don’t repeat my mistakes.

Computer is not one AI model. It’s an orchestration layer on top of 19 models. When you write a prompt, a routing system reads it, breaks it into subtasks, creates a separate sub-agent for each one, picks which model each agent runs on, and executes all of them.

Vague prompts cost real money. “Research the AI agent market and put together a report” might spawn four sub-agents. One scanning news. One pulling funding data. One analyzing strategies. One formatting. Each agent picks its own model. None of them asked whether you needed all four workstreams.

So how do you control a system that decides on its own how many agents to spin up?

I use a mental checklist before writing any Computer prompt. Five questions. I call it AGENT, but it’s a planning tool for YOU, not a formatting standard for Computer.

Computer reads natural language. It doesn’t need labeled sections. What it needs is clarity and constraints.

Before you write any prompt, ask yourself these five questions.

A: Assignment. What’s the one deliverable? Write the end goal of the task in a sentence.

G: Guardrails. What should Computer NOT do? This is the highest-leverage question. Every topic you don’t exclude is a rabbit hole. Every output type you don’t rule out (images, video, code) is a sub-agent. Every source you don’t cap is more credits burned.

Three things to always constrain. Scope: what to leave out and how far back in time. Agents: what types of output to suppress (no images, no video, no code if you just need text). Sources: cap the number. Ten well-chosen sources produce better results than forty scraped indiscriminately.

E: Engine. (Optional, experimental.) Can you specify which models to use? Computer lets you assign models to sub-tasks. I’ll be upfront: I haven’t confirmed this reliably overrides Computer’s default routing. The orchestrator makes its own decisions. But I mention it in some prompts because my results were more predictable with model hints than without. Try it. Check the model attribution in your results. See if it actually changes anything for your tasks.

N: Notify. When should Computer stop and check with you? This is the most important part after the deliverable itself. Always include: “Before you start, propose a plan and wait for my confirmation.” That one line forces Computer to show you what it’s about to do before it spends a single credit. You catch over-engineering before it costs anything.

T: Target. What does the output look like? Format, length, audience, delivery. Always set a length cap. Without one, synthesis agents keep writing until they exhaust the material. Always specify delivery. Computer connects to 400+ integrations and might try to send your output somewhere you didn’t intend.

Then write your prompt as a natural paragraph. Include all five elements. No labels needed. Computer reads English.

Here are three before-and-afters showing this in practice.

Before:

Research the AI agent market and create a comprehensive report 
on the major players, their pricing, features, and market position.

This prompt gave Computer no constraints. It created 6+ sub-agents, pulled sources on AI agents broadly (not just the market segment I wanted), generated comparison charts I never asked for, and spent credits on a “future predictions” section that added zero value.

After:

Build a comparison of the top 5 AI agent platforms (Perplexity 
Computer, Claude Cowork, OpenClaw, Manus, Operator) covering 
pricing, setup effort, and research quality. March 2026 data only. 
Exclude developer-only frameworks like LangChain, CrewAI, AutoGen. 
Maximum 8 sources per platform. No images, video, or code generation. 
Deliver as a markdown document with a comparison table at the top 
and a 300-word section per platform. 2,000 words max. Before you 
start, propose a task plan and wait for my confirmation.

Computer proposed a clean 3-phase plan. I removed one unnecessary research phase before it started. The output was tighter, more focused, and significantly cheaper.

Before:

Set up a system that monitors AI news and sends me a daily 
summary of what matters.

Computer created a monitoring agent, a summarization agent, a formatting agent, and a delivery agent. It pulled from dozens of sources per run and sent me everything, including noise I would have filtered out myself.

After:

Scan TechCrunch AI, The Information, Semafor Tech, VentureBeat AI, 
and Ars Technica AI for product launches and funding rounds from the 
past 24 hours. Only include product launches and funding rounds. 
Exclude opinion pieces, analysis, tutorials, and policy news. 
Maximum 5 items. Give me a numbered list with one sentence per item 
and a source link. Under 300 words. Only notify me if at least 2 
qualifying items are found. If fewer than 2, stay quiet.

The daily output went from a 1,500-word dump to a focused 200-word briefing. More useful. Far cheaper.

Before:

Build me a dashboard that shows my team's project status.

Computer interpreted “dashboard” as a full-stack application. It scaffolded a React frontend, wrote API routes, attempted to set up a database, and tried to deploy to Vercel. I wanted a single static HTML page.

After:

Build a single-file HTML project status dashboard showing 6 projects 
with name, owner, status (on track / at risk / blocked), and last 
updated date. Single HTML file only with inline CSS and JS. No 
backend, no API calls, no database, no deployment. No images or 
video. Dark mode, responsive. Exactly 6 project cards with those 
4 fields. Nothing else. Propose your approach before coding and 
wait for my confirmation. Do not deploy anywhere.

Computer built exactly what I asked for in one pass. No unnecessary infrastructure.

The pattern across all three. Same task. The specific version tells Computer exactly what to build, what NOT to build, and when to check with you before spending.

Three Mistakes That Drain Credits

Using Computer for tasks a cheaper tier handles. Computer spins up agents, models, and a full sandbox for every task. If Pro Search can answer it in 10 seconds, don’t waste Computer credits.

Not watching for retry loops. When Computer hits a failed connector, it retries. Then retries again. No hard stop. A broken step can silently drain hundreds of credits. Stay present for first runs. Automate after the workflow is proven.

Not calibrating before you scale. Run a lightweight version first. 3 sources instead of 10. Same structure, smaller scope. That tells you the full cost before you commit.

Quick Credit-Saving Tips

These are habits that save credits every day. Small things that add up fast.

Add a check-in step in plain English. Append this to any prompt: “Look, before you build anything, just double check with me what that looks like.” That one sentence prevents Computer from building things you didn’t actually want.

Be specific about what you want. “Research AI agents” is expensive. “Compare pricing of these 5 specific AI agents from March 2026” is cheap. Specificity reduces the number of sub-agents the orchestrator spawns.

Pay attention to system warnings. Computer has a built-in safeguard. When you give it a massive objective, the system pauses, warns you about credit consumption, and asks for approval. Don’t just click yes. Read the warning. Evaluate whether the scope is what you intended.

Keep your Custom Instructions short. You get 1,500 characters. Don’t use all of them. Unnecessary text increases your context window, which burns credits faster on every task.

Watch multi-chain tasks closely. Multi-chain prompting, app building, and long research sessions eat credits much faster than you’d expect. Monitor at perplexity.ai/account/usage.

3. Honest Limitations and Structural Risk

I need to be straight with you about some things nobody else is discussing.

It’s not a search engine. Treat it like one and you’ll burn credits on results Pro Search delivers in 10 seconds.

Connector reliability is uneven. Some work flawlessly. Others are flaky. OAuth tokens expire. The marketing says 400+. The number of battle-tested ones is smaller. Test before you automate.

Credit transparency is a real problem. No per-task pricing table. You find out the cost after. For a consumption-based product, that’s a significant gap.

And then there’s the structural risk. I wrote a full thread about this on X. Computer’s core reasoning runs on Claude Opus 4.6, built by Anthropic. Deep research runs on Gemini, built by Google. Lightweight tasks go to Grok, built by xAI. Long-context recall uses ChatGPT 5.3, built by OpenAI.

Perplexity built none of them. The product is a routing layer.

Every model provider is already building orchestration in-house. Anthropic ships Claude Code and Cowork. OpenAI has Operator. Google has Gemini with native tool use. The moment these models get good enough at everything, the routing layer becomes a feature someone else bundles for free.

Don’t sleep on it. It’s like OpenClaw for non-technical folks. And it’s buoying Perplexity’s growth. They’re back on Ramp’s list of fastest-growing software vendors for businesses.

[Bonus] Downloadable Takeaway

That’s it for today’s deep dive.

I wrote an extended version for paid members with seven more Computer workflows for PMs. Read it on Product Growth →

I watched Mike Krieger’s (former CPO at Anthropic, current leader of Labs) conversation with Dan Shipper and it was fascinating. Sharing my takeaways:

1/ The hard part of building flipped overnight. AI can go from zero to feature-complete in hours. That’s not the bottleneck anymore. The bottleneck is knowing what to cut. Claude will happily build you a monstrosity of overbuilt functionality if you let it.

2/ Stop building indoor trees. Krieger’s metaphor is perfect. Vibe coding lets you build an entire product indoors without ever exposing it to the wind of real users. If you’re 6 weeks into a build and nobody outside your team has touched it, you’re growing an indoor tree.

3/ The next generation of software is agent-native. Agent-native means the model can navigate and modify the product’s own primitives. Claude Code already knows about its own skills and can modify its own harness.

4/ Small teams are about to eat large teams alive. The unit of value is shifting from coordinated teams to autonomous individuals. Designers writing as much code as engineers. Scaling headcount too fast is now a net negative because models change so quickly that half your codebase might need to be deleted every few months.

The winning role is the GM who spikes on product sense and uses AI to manage the entire stack.

Watch Full Conversation

That’s all for today. See you next week,

Aakash

Share

P.S. Want my AI tool stack? Join my bundle. Want my job searching coaching? Apply to my cohort.

Everyone covered the ML side. But the pattern underneath works on anything you can measure. Skills, prompts, agents, email, page speed. I’ve spent the last two weeks pulling it apart, and today’s deep dive is your complete guide to using it.

But first, I cover Google Maps getting its biggest AI upgrade in over a decade, plus all the news that matters from the week.

Here’s what you need to know:

Ariso: Your AI Chief of Staff

I hired a chief of staff this year. It costs $0/month and lives in my Slack.

Ariso’s AI connects to my email, calendar, and docs, then handles meeting prep, follow-ups, and the busywork I used to lose hours to. It remembers every conversation and gets smarter the more I use it.

Built by cybersecurity founders, so your data is encrypted per-user and never trains their models.

I personally recommend this tool. Try it free at ariso.ai/aakash.

Try Ariso

Top News: Google Maps Just Got Its Biggest Upgrade in Over a Decade

OpenAI and Anthropic are building the best AI chatbots in the world. But Google just reminded everyone it owns the browser, the email, the maps, and the search engine.

Google announced two major features last week. Ask Maps is a Gemini-powered conversational layer that lets you type things like “My phone is dying, where can I charge it without having to wait in a long line for coffee?” and get real answers pulled from 300 million+ places and 500 million reviews. The second feature, Immersive Navigation, replaces the flat map with a full 3D view of buildings, terrain, stop signs, and lane markers. It analyzes Street View imagery and aerial photos using Gemini models to give you a real picture of what’s actually outside your windshield.

Here’s why this matters for builders. Google isn’t just shipping AI features in a chatbot. It’s embedding Gemini into the products 2 billion people already use every day. Maps, Gmail, Chrome, Android, Search. That’s not a model play. That’s a distribution play. And right now, no one else has that combination.

OpenAI can answer questions about the world. Google can answer questions about you.

The Other News That Mattered

• Travis Kalanick spent eight years building a ghost kitchen company. That was the cover story. He just renamed it Atoms and revealed he’s been building a robotics company across food, mining, and autonomous transport. Thousands of employees, 110 cities, nobody was allowed to put the company name on LinkedIn. The same board that pushed him out of Uber is now reportedly backing him.

• A tech entrepreneur in Sydney used ChatGPT to help design a custom mRNA cancer vaccine for his dying dog. Paid $3,000 to sequence the tumor DNA, ran mutations through AlphaFold, produced a personalized vaccine in under two months. The tumor shrank 75% while Moderna is spending billions to scale the same workflow for humans.

• Anthropic shipped Dispatch, a new Cowork feature that lets you control your desktop AI agent from your phone. Pair your phone with Claude Desktop, message from anywhere, come back to finished work. I’ve been running Cowork since launch and this was the missing piece.

• An engineer in Morocco open-sourced a phased array radar you can build from GitHub files. Commercial systems start at $250,000. The AERIS-10 does electronic beam steering and multi-target tracking at 20km range for roughly $5,000-$15,000 in parts. Open source collapsing defense contractor paywalls in real time.

Resources

• Claude now builds interactive charts, diagrams, and visualizations right inside your conversation. Not in a side panel. Not as a downloadable artifact. Ask it to explain compound interest and you get an interactive curve you can play with. This is the kind of feature that makes you stop opening Excalidraw for quick explanations.

• If you’re building Claude skills, I urge you to make them self-improving. Not just agents with skills, but agents with skills that get better over time. SKILL.md is clearly here to stay. The question now is whether your skills are static instructions or living systems that learn from their own outputs.

• I urge you to start using subagents. Codex just dropped subagent support. Instead of running everything in one thread, you can now spin up parallel agents that handle different parts of a task independently. If you’re not breaking complex tasks into coordinated subtasks yet, you’re leaving speed on the table.

Tools

• Google Stitch just got a massive update aimed directly at Figma’s throat. Google acquired Galileo AI, killed the $39/month subscription, and now gives you 350 design generations per month for free. The DESIGN.md export captures your entire design system as agent-readable markdown that flows into Cursor, Claude Code, and Firebase. Stitch is Maps for the design workflow.

• Claude Code now has Remote Control. Kick off a task in your terminal, pick it up from your phone while you take a walk. Claude keeps running on your machine and you control the session from the Claude app or claude.ai/code.

Fundraising

• Cursor is raising at a $50 billion valuation. In the same week, Elon pulled two of its product leaders into xAI to build the coding product missing from Grok’s stack. AI coding is now a $5B+ market and every frontier lab has a product printing money. xAI was the only one that didn’t.

The Ultimate Autoresearch Guide

Subscribe now

You’ve been settling. Your ad copy converts at 2%. Your onboarding emails get opened but nobody activates. Your AI-generated content needs three rewrites before it’s usable. You know all of this could be better, but testing 50 variations by hand isn’t a real option when you’re also trying to run a business.

Andrej Karpathy built the fix. He called it autoresearch, and 42,000 people starred it on GitHub in its first week. Fortune named it “The Karpathy Loop.”

Every article about it focused on the machine learning angle. They all missed the bigger story. The pattern underneath works on ad copy, email sequences, video scripts, job posts, and anything else where you can define what “good” looks like. I spent two weeks pulling it apart, going through Karpathy’s repo, the community forks, and the real-world applications people are building on top of it. This guide shows you how to use it on the things you actually make every day.

You don’t need a GPU. You don’t need to know anything about ML.

The tech world moved on after the ML headlines. That was premature. The core loop has nothing to do with training neural networks. It takes whatever you’re producing with AI, runs 100 variations while you sleep, keeps the ones that score better, throws away the ones that don’t, and saves the improved version for when you wake up.

1. What autoresearch actually does

Karpathy builds small language models for fun. The workflow looks the same whether you’re training a model or tuning a prompt: make a change, test it, check if the output got better, keep or discard, repeat. On a good day he’d get through 8-10 cycles. Most of that time was spent waiting.

He automated the entire thing.

Three files power the system. One file the agent edits (the thing being optimized). One file the agent can never touch (the scoring criteria, because if it could, it would just make the test easier). And one instruction file written by the human that tells the agent what to try and how to behave.

Each cycle takes about 5 minutes. The agent reads the current version, decides what to change, makes the edit, runs the test, and checks a single number. Score went up? The change gets committed to git. Score went down or stayed flat? git reset wipes it clean. Next round starts.

Twelve cycles per hour. A hundred overnight. About $25 in compute.

Karpathy ran it for two days on code he’d already optimized by hand for months. The agent found 20 improvements he’d missed, including a bug in his attention implementation. All 20 stacked and transferred to a larger model. 11% faster.

Tobi Lutke, the Shopify CEO, tried it the same night. 37 experiments while he slept. His smaller model started outperforming his larger one. Then he turned the loop on Liquid, Shopify’s templating engine, and got 53% faster rendering from 93 automated commits.

The ML part was the proof of concept. The loop itself is the product.

2. Why this works beyond ML

Forget the neural network part. The loop runs on any output where three conditions hold.

You can score the output with a number. Not a feeling, not a vibe check. Binary questions work best. “Does the ad headline mention a specific pain point?” Yes or no. Run 30 outputs through 5 binary questions and you get a percentage. That percentage is your metric.

The scoring runs without a human. Your coding agent builds an evaluation script that generates outputs, grades them against your questions, and prints the score. No one needs to sit and read 30 ad variants at 2am. The machine scores them and moves on.

Only one file changes per round. Your prompt, your template, your skill file. The agent edits this single file. Everything else is locked. This constraint is what makes the loop work. If the agent could edit the scoring criteria too, it would just make the test easier instead of making the output better.

All three true? The loop works. Any one missing? It won’t. That’s the whole filter.

Your job is choosing what “better” means and writing the yes/no questions that define it. The agent handles the 50 rounds of iteration you were never going to do manually. I run this in Claude Code, but Cursor, Windsurf, Codex, and Antigravity all work. Any coding agent that can read files, edit files, and use git.

3. How to set it up

Three steps. Five minutes of setup, then the agent takes over.

Step 1: Clone the repo

git clone https://github.com/karpathy/autoresearch ./autoresearch-reference
git init && git add . && git commit -m "setup"

This gives your coding agent a reference copy of Karpathy’s convention. If you already have a prompt or skill file you want to improve, drop it in the same folder.

Step 2: Write your eval criteria

This is the human’s job and the part where most people succeed or fail.

You’re writing yes/no questions that describe what a good output looks like. Here’s what I used for a landing page copy skill:

1. Does the headline include a specific number or measurable result?

2. Is the copy completely free of buzzwords?

3. Does the CTA use a specific action verb tied to the product outcome?

4. Does the first sentence after the headline name a specific pain point?

5. Is the total copy between 80 and 150 words?

Five binary questions. The agent generates 30 outputs per round, scores every one, and produces a single percentage.

Don’t know where to start? Just describe what goes wrong when your current output fails. “The headlines are generic, there’s too much jargon, the CTA is weak.” Claude Code will convert those complaints into binary scoring criteria.

Three to six questions is the range. Below three and the agent finds loopholes. Above six and it starts gaming the checklist instead of improving what the output actually sounds like.

Step 3: Tell Claude Code to run the loop

Read autoresearch-reference/program.md to understand Karpathy's 
autoresearch convention. I want to optimize [my skill / prompt / 
system prompt] using these scoring criteria: [your yes/no questions]. 
Test against these inputs: [2-3 realistic scenarios]. 

Set up the evaluation harness, run the baseline, then enter the 
autoresearch loop. One change per round. Keep if score improves, 
revert if it doesn't. Never modify the evaluation after creating it.

One prompt. Claude Code reads the convention, builds the eval script, generates a baseline score, and starts iterating.

4. Six use cases you can run tonight

Use case 1 - Self-improving skills

You need to steal this autoresearch skill optimization use case immediately.

I ran this on a landing page copy skill with the 5 eval criteria. The baseline came back at 41.3%

Round 1: Claude Code saw that “specific number in headline” was failing 80% of the time. It added a rule to the skill requiring specific numbers. Score jumped to 68%. Kept.

Round 2: Buzzwords were the next biggest failure. It added a banned words list. 79%. Kept

Round 3: CTAs were still generic. It added a worked example showing what a strong CTA looks like. 90%.

Round 4: It tried tightening the word count. Score dropped to 82% because shorter copy hurt the CTA.

41% to 92%. 4 rounds. Three changes kept, one reverted. The eval caught the regression automatically.

Start with the skill that frustrates you most. The one where you re-run 3 times before getting usable output.

Use case 2 - Cold email that optimizes itself

Your outreach reply rate is at 2-3%. You’ve tried manual A/B tests but it takes weeks to get signal.

Eval criteria: Under 75 words? References the prospect’s specific role? Ends with a concrete question? First two sentences include a specific number or result?

The agent discovers what human copywriters learn over years: shorter beats longer, specificity beats vague, risk-reversal beats feature lists. Eric Siu, founder of ad agency Single Grain, is already building this for his clients. His framework replaces the training script with the email itself. The agent modifies the subject line or CTA, measures reply rate, keeps or discards. Siu’s math: most marketing teams run ~30 experiments a year. This loop runs 36,500+.

MindStudio documented teams using this pattern seeing reply rates move from a 2-4% baseline to 8-12% within four to six weeks. For the live version, connect to your email platform’s API (Instantly, Apollo, Lemlist) and deploy to GitHub Actions on a cron schedule.

Use case 3 - Ad copy that tests itself

You’re running Meta or Google ads and testing 2-3 headline variants per week by hand. That’s maybe 150 tests per year. Autoresearch runs 100 overnight.

Eval criteria: Is the headline under 40 characters? Does it name a specific pain point in the first 5 words? Does the body mention a number or timeframe? Is the CTA a verb phrase, not a noun phrase (”Start saving” not “Savings plan”)? Does it avoid every word on your brand’s banned list?

Siu’s framework at Single Grain applies here directly. The agent iterates on your ad copy template the same way it iterates on a training script. Each round generates 30 ad variants, scores them all, keeps the template change if the average score went up, reverts if it didn’t. MindStudio’s guide walks through the full implementation for landing pages and ad creatives, including how to connect the loop to your CMS via API. You wake up to a template that produces compliant, specific, action-oriented copy on the first try.

Use case 4 - Short-form video scripts

Your Reels and TikToks hit when the hook lands and miss when it doesn’t. You know this intuitively but you test hooks by posting and waiting 48 hours for the algorithm to tell you.

Eval criteria: Does the opening line create a specific curiosity gap? Is there a surprising claim in the first 2 sentences? Does it follow a single narrative thread (not 3 loosely connected points)? Is the script under 60 seconds when read aloud? Does the last line set up the next piece of content?

The agent can’t predict virality. But it can enforce the structural patterns that correlate with watch time. witcheer, Head of Growth at crypto platform Yari Finance, ran autoresearch overnight on a Mac Mini M4 and found that “the model got better by getting simpler.” The same principle applies to scripts: the agent strips filler, tightens hooks, and enforces a single thread. The creative ideas are still yours. The structural consistency is the machine’s job.

Use case 5 - Onboarding email sequences

You wrote your 5-email onboarding sequence six months ago. Open rates are fine. But activation is stuck at 30% and you’ve been too busy to rewrite them.

Eval criteria: Does each email have exactly one CTA? Is the email under 100 words? Does the subject line include a specific benefit or number? Does the first sentence reference the user’s last action (not a generic greeting)? Does it avoid “we” in the first paragraph (focus on them, not you)?

One file changes: your email template. The eval generates 30 onboarding sequences per round, scores every email in every sequence, and gives you an aggregate. If you want to go further, 35 agents on the Hyperspace network ran 333 experiments in a single night, completely unsupervised. Agents on cheap hardware found optimizations the expensive setups missed, because constraints forced creative solutions. The same distributed approach works on email: spin up multiple agents testing different parts of the sequence in parallel.

Use case 6 - Job descriptions

Most job posts are written once and never optimized. They’re too long, full of jargon, and read like a legal document. Then you wonder why qualified people don’t apply.

Eval criteria: Is it under 400 words? Does it name a specific project the hire will work on in the first 90 days? Is it free of internal jargon and acronyms? Does it include compensation range? Does the first sentence describe impact, not responsibilities?

This is one of the cleanest autoresearch applications because the eval criteria are obvious and binary. The agent strips jargon, forces specificity, and front-loads impact. After a few rounds you have a template that produces job posts people actually want to read. Forward this to your recruiting team and watch application quality change.

5. What most people get wrong

After spending two weeks with this, I’ve seen the same three mistakes kill autoresearch runs before they start.

Mistake 1: Vague eval criteria.

“Is the copy compelling?” is not a question an agent can score. “Does the headline contain a specific number?” is. Every criterion must be binary. Pass or fail, yes or no, no sliding scales. The moment you introduce a 1-7 rating, the agent starts optimizing for 4s that technically score well but read like garbage. If you can’t explain to a stranger how to score it in one sentence, rewrite the criterion.

Mistake 2: Touching the eval mid-loop.

Your baseline comes back at 35% and your first instinct is to soften the criteria so the number looks better. This is the single fastest way to waste a run. The eval is locked the moment the first round starts. If you change the scoring mid-loop, every previous round’s data becomes meaningless. You’re comparing apples to oranges across your own experiment log. If the eval is wrong, stop the run, fix it, and restart from scratch. Karpathy made prepare.py read-only for exactly this reason.

Mistake 3: Too many criteria.

Seven or eight eval questions feels thorough. In practice, the agent starts gaming the checklist. It finds combinations of changes that technically pass 7 out of 8 criteria but produce outputs that don’t read like something a human would write. Three to six criteria is the sweet spot. Enough to prevent shortcuts, few enough that the agent optimizes for real quality instead of checklist compliance. When in doubt, start with 3 and add one at a time if the output has a failure mode you didn’t cover.

The pattern is simple. The eval is the hard part. Get that right and the loop does the rest. I explained the broader vision here.

Pick one thing you’ve been meaning to improve. Write 3-5 yes/no questions that define what “good” looks like. Let the loop run tonight. Wake up to something that works.

That’s it for today’s deep dive.

I wrote an extended version for PMs with six more use cases (page speed, agent system prompts, production agents, content feedback loops, and more) plus the experiment log strategy that makes this compound over time. Read it on Product Growth →

I watched David Senra’s conversation with Marc Andreessen and it’s one of the best founder interviews I’ve heard this year. Sharing my takeaways because this one hit different.

1/ Great founders have near-zero introspection. Andreessen says the most successful entrepreneurs spend almost no time self-analyzing or dwelling on the past. Sam Walton had zero inner dialogue about his feelings. He just woke up every day and kept building Walmarts. The whole concept of sitting around analyzing yourself is newer than we think. Before Freud, nobody did it. They just went out and built things.

2/ Founders beat professional managers every time during rapid change. This is a core a16z thesis. Formally trained managers can maintain status-quo operations fine. But when industries undergo massive technological shifts, they fail. It is much more effective to start with an innovative founder and teach them how to manage than to expect a professional manager to invent like a founder. If you’re a PM watching AI reshape your industry, this should feel personal.

3/ Every major technology faces a moral panic. Every single one. The bicycle got doctors warning about “bicycle face.” The printing press, jazz, TV, video games, social media. Same pattern. The world is far more malleable than people realize, and founders who push past the panic with maximum effort can force the world to recalibrate around their ideas. AI is getting the same treatment right now.

4/ Elon Musk runs a management style nobody else is willing to try. Traditional CEOs sit behind layers of management that filter the truth. Andreessen compares it to IBM’s historical “big gray cloud.” Musk does the opposite. Every week he maps the exact production bottleneck across each of his companies, bypasses every management layer, and works hands-on with frontline engineers to fix problems directly. That speed is the competitive advantage, not the rockets or the cars.

Watch Full Conversation

That’s all for today. See you next week,

Aakash

Share

Karpathy just open-sourced a research lab that runs while you sleep. Netflix dropped $600M on a 16-person startup. And human brain cells learned to play Doom.

In today’s issue, I cover all the news that matters from the week.

And for today’s deep dive, I go deep on the SOTA Image Gen model that dropped 2 weeks ago, Nano Banana 2:

Nimbalyst: The Visual Workspace for Claude Code and Codex

Today, you describe what you want in a terminal, the agent writes files to disk, and you open them somewhere else to see what happened. Like giving renovation instructions through text messages.

Now imagine working in the same visual workspace as your agent. Planning, coding, diagramming, tracking tasks, all connected. You see what the agent is doing and approve it. The agent sees what you’re doing and uses it.

This is Nimbalyst. Download it for free here:

Try Nimbalyst

There’s a billion AI news articles every week. Here’s what actually mattered.

The Week’s Top News: Karpathy Just Turned One GPU Into a Research Lab

Andrej Karpathy dropped autoresearch last weekend, and the internet went sideways. The repo crossed 27,000 stars within days.

Here’s what it actually is: A 630-line Python script that puts an AI agent in charge of running ML experiments on a single GPU:

• The human writes a Markdown file with high-level instructions.

• The agent modifies the training code, runs a 5-minute experiment, checks if the result improved, commits the good ones, and loops back. Twelve experiments per hour. Around 100 while you sleep. No breaks, no forgetting, no lab politics.

This matters because it changes who can do ML research. A PhD student at a mid-tier university with one GPU can now run 100 experiments overnight. A startup that can’t afford a research team gets automated ablation studies across hyperparameters, architectures, and training schedules while the founders sleep. A solo developer testing a new loss function gets 12 runs per hour instead of doing one, checking Slack, doing another, losing the thread.

The use cases go beyond academic papers. Product teams can A/B test model variants against real evaluation criteria. Infrastructure teams can automate regression testing across model versions. Fine-tuning shops can let the agent grind through LoRA configs until it finds the one that actually moves the benchmark.

Right now it’s one agent on one GPU. Karpathy already outlined what comes next: hundreds of agents running parallel research branches, reading each other’s results, combining findings.

A 630-line script just turned every GPU on the planet into a potential research lab.

The Other News That Mattered

• Netflix acquired Ben Affleck’s AI startup InterPositive for a reported $600M, one week after walking away from an $83B Warner Bros deal. InterPositive trains models on a production’s own dailies and lets filmmakers fix continuity errors, relight shots, and remove wires in post. Netflix spends $20B on content annually. Even a 10% cut in post costs pays for this many times over.

• Cortical Labs grew 200,000 human neurons on a chip and taught them to play Doom. An independent developer with zero biotech experience built the integration in a week using a Python API. Pong took 18 months. Doom took a week. Your brain runs on 20 watts. A GPU cluster draws megawatts. The energy economics here are not close.

• Claude Code launched local scheduled tasks via a new /loop command. You describe a task, set a cadence, and Claude runs it in the background while your computer stays on. Check error logs every few hours, create PRs for fixable bugs, generate morning Slack summaries.

• ChatGPT now has 140 million people using it weekly for math and science, and they just added interactive visual modules for 70+ topics. Adjust a variable in Ohm’s law, watch the graph update in real time. Khan Academy spent 18 years reaching 2 million AI tutor users. OpenAI has 70x that and just started adding the interactive layer.

Resources

• Within a year, every company over 50 people will have at least one full-time Agent Builder. This piece on the rise of the Agent Builder is the clearest breakdown I’ve seen of what that role actually looks like.

• If your lawyer isn’t using AI like this, find a new lawyer. Zack Shapiro’s piece on running a Claude-native law firm is one of the best real-world examples of AI replacing entire workflows, not just speeding them up.

• OpenAI’s newer models hallucinate 3x more than the ones they replaced. o1 hallucinated 16% of the time. o3 hit 33%. o4-mini hit 48%. They published a paper explaining exactly why they can’t stop it and the answer is uncomfortable: the system that produces intelligence and the system that produces hallucinations are the same system.

Tools

• Google’s new Workspace CLI is the fix nobody knew they needed. One developer found his Google MCP setup loading 37,000 tokens into context before the agent even started working. The CLI sidesteps that entirely. Same capabilities, fraction of the overhead. 100+ agent skills. As agents get smarter, the bottleneck shifts from “can it access the tool” to “how much context does it cost.” CLIs win that math every time.

Fundraising

• Fei-Fei Li raised $1B for World Labs. Now Yann LeCun raised $1.03B for AMI Labs, a company that is 4 months old with zero product and zero revenue. Both closed within three weeks with overlapping investors. The bet they’re both making: large language models have hit a ceiling, and the future belongs to world models that understand actual physics. If they’re right, a lot of companies building physical-world applications on top of current LLMs picked the wrong foundation.

The Ultimate Guide to Nano Banana 2 🍌

Nano Banana 2 dropped two weeks ago and I immediately started testing.

What I found genuinely surprised me. This isn’t a cheaper version of Pro (the last drop in November). It’s a better model:

• PDF input up to 50MB

• Google Image Search grounding

• Real-time web search grounding

• A context window double the size of Pro

• A 512px resolution tier that makes high-volume economics work in a way they never did before

• And a thinking mode that reasons through compositions before generating - which changes first-pass quality more than I expected

I ran it through two weeks of real work: competitive analysis, prototype iteration, deck prep, stakeholder docs, UGC seeding.

Here’s what I learned, and every prompt worth stealing.

Subscribe now

How to Access Nano Banana 2

30 seconds to set up.

1. Go to gemini.google.com

2. Click on “Tools” and then “🍌 Create image”

3. Switch to the Thinking model

   

   That’s it. You’re on Nano Banana 2.

The 13 Most Useful Workflows

I tried out 100+ things and found the 13 you should care about:

1/ The Frankenstein Mockup

You’re walking into a product review and you need to show what a new feature could look like. Design is tied up for days even if you are using AI. You have your app’s nav in one tab, a competitor layout you like in another, your brand color palette somewhere, and a whiteboard sketch from last Tuesday.

Attach all four. That’s the entire prompt setup.

Using the 4 attached reference images — [1] my app's current navigation bar, 
[2] a competitor's card layout I want to adapt, [3] our brand color palette 
screenshot, [4] my whiteboard sketch of the new feature — generate a high-fidelity 
UI mockup of [FEATURE NAME]. 

Match the component style from image 1, adapt the card structure from image 2, 
apply the exact colors from image 3, and reflect the rough layout from image 4.

Render as a mobile screen, 9:16 aspect ratio, white background, clean and 
implementation-feasible. Use realistic UI elements, no placeholder icons.

2/ Research Insight to Shareable Infographic

You’ve done the user research. Nobody reads the 40-page report. You need something you can drop in Slack, paste into a deck, or print for a workshop before the next standup.

Nano Banana 2’s text rendering is the reason this works cleanly now. Earlier versions of the model couldn’t hold sharp legible type across the whole image. This one does.

Generate a clean, professional infographic summarising the following 
research insight:

[PASTE YOUR KEY FINDING HERE — 2-3 sentences with the core stat]

Layout:
- Top: Bold headline stating the core finding
- Middle: 3-4 data points in labelled visual blocks, each with a large 
  typographic number as the anchor
- Bottom: One "what this means for us" callout in a distinct colour band

White background, [BRAND COLOR] as primary accent, clean sans-serif 
typography, no stock photo elements, data visualisation only. 
Aspect ratio 4:5, 2K resolution.

3/ The Competitive Visual Grid

You need to show leadership how three competitors are handling the same problem. Building a real comparison takes a designer half a day and your deck is in two hours.

Describe what you need. Let Nano Banana 2 search for the current UI and render the comparison.

Search the web for the current [SPECIFIC SCREEN — e.g., "onboarding empty 
state" / "checkout flow" / "settings page"] of [COMPETITOR 1], [COMPETITOR 2], 
and [COMPETITOR 3].

Then generate a clean side-by-side comparison grid showing how each 
product handles [SPECIFIC UX PATTERN].

Format as a 3-column grid with each competitor's name as a bold column 
header. Under each, show a simplified but realistic mockup of their approach 
with 2-3 annotation callouts highlighting the key design decisions.

White background, 16:9 aspect ratio, minimal styling, annotation callouts 
in small grey labels with arrows. This is for an internal strategy presentation.

4/ PRD to Visual

You’ve written a solid spec. Nobody reads it end-to-end. You spend the first 20 minutes of every product review re-explaining the concept before the actual discussion starts.

Feed the PDF. Let it generate the visual summary instead of you rebuilding the explanation every time.

I'm attaching my product requirements document as a PDF. Read it fully, 
then generate a clean visual overview of [FEATURE NAME] that shows:

1. The user problem being solved (before/after comparison)
2. The proposed solution as a labelled product diagram
3. The key user journey in 3-4 annotated steps

White background, [BRAND COLOR] accents, bold sans-serif headers, 
clean connecting arrows between elements.

Design this to be readable at 1080px width without zooming — 
it'll be embedded in Notion or shared in Slack.

5/ The Surgical Fix

You generated a mockup that’s 80% right. One element is wrong. You don’t want to start over.

This is what semantic masking is for. Describe the region by what it contains, not by coordinates. Nano Banana 2 changes only that and leaves the rest exactly as it was.

Keep the entire image exactly as it is. Make only this one change:

[SPECIFIC EDIT — e.g., "Replace the blue primary button in the bottom 
right with a button using hex color [YOUR BRAND HEX]. Change the button 
label to '[NEW LABEL]'. Keep the size, position, and shape identical."]

Do not change the background, layout, typography, icon placement, 
or any other element. Only modify what I described above.

6/ App Store Lifestyle Screenshots

App store screenshots that show the product being used by a real person in a real context outperform clean product shots. A phone on a white background tells someone what your app looks like. A person using your app at a coffee shop tells them what their life looks like with it.

Most teams don’t do lifestyle screenshots because they require a photoshoot. With Nano Banana 2, you skip the photoshoot entirely.

Generate a realistic lifestyle App Store screenshot showing 
[YOUR APP NAME] being used in an everyday context.

The person: [AGE, GENDER, DESCRIPTION — e.g., "a woman in her 
early 30s, calm, present, wearing casual home clothes"]

The setting: [ENVIRONMENT — e.g., "sitting cross-legged on a sofa 
by a large window, morning light, ceramic mug nearby"]

Camera angle: the person is looking DOWN at their phone naturally. 
The phone is NOT held up toward the camera — it sits at a natural 
angle in their hands or lap. Frame from eye level so the face is 
the clear primary subject. The phone screen should be readable 
but secondary to the human.

The screen shows: [DESCRIBE YOUR KEY UI — e.g., "a dark navy 
meditation screen with a soft glowing circle and 'Breathe in...' text"]

Shooting style: natural light, f/1.8 shallow depth of field, 
face in sharp focus, background soft. Editorial feel, not a product shoot.

Output at 9:16, 2K resolution.

7/ Feature Launch Social Asset

You just shipped something. You want a LinkedIn post that stops the scroll. You don’t want another screenshot with a gradient background. You want a person — your person, your user — in the moment your feature solves their problem.

Generate a realistic photo-style image for a LinkedIn feature launch post.

The scene: [DESCRIBE THE HUMAN MOMENT — e.g., "a product manager in 
her early 30s, sitting at a standing desk in a modern open office, 
looking at her laptop with a visible expression of relief — she just 
found the thing she was looking for"]

Demographic: [AGE RANGE, GENDER, ETHNICITY — e.g., "South Asian woman, 
late 20s to mid 30s, professional but not overly formal"]

Setting: [CONTEXT — e.g., "daytime, natural light from a nearby window, 
laptop open, coffee cup on the desk, slightly cluttered workspace — 
looks real, not staged"]

The laptop screen should show [YOUR PRODUCT SCREEN — attach a screenshot 
or describe the key UI moment]. It should be readable but not the focus 
— the human reaction is the focus.

Shooting style: candid documentary feel, slightly shallow depth of field, 
natural colour grading. Looks like a real photo a colleague took, 
not a commercial shoot.

Output at 4:5 (LinkedIn optimal), 2K resolution.

8/ Localized GTM Assets in One Prompt

You’re launching in India, the UAE, and Germany simultaneously. The old way means three separate briefings, a localization vendor, two weeks of back and forth, and assets that still don’t quite feel right because the translator and the designer never spoke to each other.

Nano Banana 2’s text rendering and multilingual support changes this. One prompt generates the same launch banner in Hindi, Arabic, and German — correct typography, right-to-left layout for Arabic, cultural context baked in. What used to be a two-week vendor process is a single generation.

Generate 3 versions of a launch banner for [PRODUCT/FEATURE NAME].

All 3 versions should be visually identical in layout and design, 
but rendered in different languages with culturally appropriate typography:

Version 1: Hindi — Devanagari script, warm and celebratory tone
Version 2: Arabic — right-to-left layout, Arabic script, 
           ensure the design mirrors correctly for RTL reading
Version 3: German — clean sans-serif, precise and professional tone

Banner layout:
- Top: Product logo placeholder
- Center: Hero headline "[YOUR LAUNCH MESSAGE]" in each language
- Below: Subheading "[YOUR VALUE PROP — 1 line]" in each language  
- Bottom: CTA button "[YOUR CTA TEXT]" in each language

Design: [PRIMARY COLOR] background, [ACCENT COLOR] for CTA, 
white text, [FONT STYLE] where Latin script is used.
Same visual hierarchy across all 3 versions.

Output each at 16:9, 2K resolution.

9/ Material and Texture for Hardware or Physical Product PMs

If you’re shipping something physical — a device, packaging, a consumer product — you know the pain of getting a photorealistic render for an investor deck or manufacturer brief without a 3D rendering team.

Nano Banana 2’s materiality framework is built for this. You’re not describing “a white box.” You’re describing “recycled kraft paper with an embossed logo, matte surface, slight texture visible under directional studio light.” It renders the difference.

Generate a photorealistic product render of [YOUR PRODUCT — e.g., 
"a wireless earbud charging case" / "a supplement bottle" / 
"a smart home device"].

Material: [EXACT MATERIALS — e.g., "matte anodized aluminum body, 
frosted glass top panel, brushed finish on the edges"]

Surface: [TEXTURE DETAIL — e.g., "subtle grain on the aluminum, 
fingerprint-resistant coating, logo embossed 0.5mm depth, centered"]

Lighting: Three-point softbox — main light upper left, soft fill 
right, rim light from behind to define edges. No harsh shadows.

Background: [e.g., "seamless warm white studio sweep" / 
"dark slate grey for premium feel"]

Composition: Product centered, angled 15 degrees, medium shot, 
full product visible. f/2.8 shallow depth of field.

Output at 1:1, 4K resolution, investor deck ready.

10/ Product Hunt Launch Hero Image

Your Product Hunt hero is the first thing 50,000 people see on launch day. Most founders put a screenshot or a gradient with their logo. The products that hit #1 almost always have a specifically composed, specifically lit hero that looks intentional.

You don’t need a designer for this anymore.

Generate a cinematic hero image for a Product Hunt launch.

Product: [NAME] — [ONE LINE DESCRIPTION]

Concept: [VISUAL IDEA — e.g., "close-up of hands typing on a laptop 
at night, screen glowing with the product dashboard, everything else 
in darkness — conveys focus and late-night momentum"]

Lighting: [STYLE — e.g., "single cold key light from the left, 
high contrast, screen glow as secondary light source — 
cinematic, not commercial"]

Camera: [LENS + ANGLE — e.g., "low angle looking up at the screen, 
wide-angle lens, slight distortion — makes the product feel significant"]

Color grade: [MOOD — e.g., "muted cool tones, slight blue shift, 
cinematic desaturation — modern and focused, not playful"]

No text in the image. Output at 16:9, 4K resolution.

11/ Product Mockup Shots Without a Photoshoot

You need a hero image for a press release, a launch tweet, or a pitch deck. No photoshoot budget. No time to brief a designer.

Describe the material. Specify the lighting rig. Define the surface. Nano Banana 2 renders a studio-quality product shot in 30 seconds that looks like it cost $3,000 to produce.

Generate a studio product shot of [YOUR PRODUCT — e.g., 
"iPhone 16 Pro showing the app's onboarding screen" / 
"MacBook Pro showing a SaaS dashboard" / 
"a physical device on a clean surface"].

Material: [e.g., "Space Black iPhone, titanium frame, matte back, 
screen showing the app's key moment"]

Lighting: Three-point softbox — [MOOD — e.g., "warm main light 
upper left, soft fill from right, subtle rim light from behind. 
Clean commercial product photography look."]

Background: [e.g., "seamless off-white studio sweep" / 
"dark textured concrete for premium feel" / 
"warm walnut wood for lifestyle feel"]

Composition: Product centered, angled 20 degrees, full product 
visible, no cropping. f/2.8, background slightly soft.

Output at 16:9, 4K resolution, press-ready.

12/ Personal Brand Headshot Style Transfer

Every founder and PM needs a headshot that doesn’t look like everyone else’s grey-gradient LinkedIn photo. The problem is getting a specific visual style — bold background, high contrast, editorial feel — applied to your actual face without booking a photographer who shoots in that style.

Nano Banana 2’s multimodal input changes this. Upload your photo. Upload a style reference. Tell it exactly what to transfer and what to leave alone.

The thing most people get wrong: they upload both images and just say “make mine look like this.” The model blends the two faces and the output looks like neither person. You need to explicitly separate what transfers from what stays.

Generate a styled headshot using the attached reference images.

Image 1 (attached): my photo — use my face, hair, and likeness exactly. 
Do NOT alter my facial features or expression.

Image 2 (attached): style reference — use the visual style only. 
Do NOT copy the face of the person in this image.

Transfer from the reference: [LIST WHAT TO TRANSFER — e.g., 
"background color and treatment, lighting style, contrast level, 
black-and-white vs color treatment, crop and framing style"]

Keep exactly: my face, my hair, my skin tone, my features, my expression.

Output at 1:1, 2K resolution.

13/ Research Insight to Styled Infographic

Your user research is dying in Google Docs. Nobody opens the synthesis deck. Nobody reads the Notion page.

The data is good. The format is the problem.

Nano Banana 2’s text rendering turns that same data into a shareable infographic in one prompt. But skip the clean corporate version the one that gets shared is the sketchnote doodle style. Hand-drawn, lined notebook paper, highlighter on the key phrases.

Here’s the prompt:

Generate a sketchnote doodle infographic in the style of a 
hand-drawn notebook page.

Background: lined notebook paper — horizontal ruled lines 
visible across the full image, warm cream paper color, 
spiral binding visible on the left edge. 
Looks like a real opened notebook.

Style: hand-drawn black ink doodle aesthetic throughout — 
imperfect lines, slightly uneven hand-lettered text, 
simple illustrative icons drawn per section. 
NOT clean vector. NOT corporate slides.

Layout: 2x3 grid of 6 equal sections with dashed border boxes. 
Banner headline at the top center in a hand-drawn ribbon shape. 
Small footer text at the bottom center.

Headline (bold hand-lettered caps inside ribbon banner):
"[YOUR HEADLINE — e.g., 6 THINGS EVERY PM SHOULD DO 
BEFORE WRITING A PRD]"

Content — 6 sections, one per box:
Section 1: [NUMBER + TITLE] — [1 LINE INSIGHT] — [DOODLE DESCRIPTION]
Section 2: [NUMBER + TITLE] — [1 LINE INSIGHT] — [DOODLE DESCRIPTION]
Section 3: [NUMBER + TITLE] — [1 LINE INSIGHT] — [DOODLE DESCRIPTION]
Section 4: [NUMBER + TITLE] — [1 LINE INSIGHT] — [DOODLE DESCRIPTION]
Section 5: [NUMBER + TITLE] — [1 LINE INSIGHT] — [DOODLE DESCRIPTION]
Section 6: [NUMBER + TITLE] — [1 LINE INSIGHT] — [DOODLE DESCRIPTION]

Highlighter accents: yellow (#FDE68A) and pink (#FBCFE8) — 
use on the 1-2 most important words per section only. 
Looks like someone went back over it with a real highlighter.

Footer in small neat handwriting at the bottom center:
"Subscribe to aibyaakash.com"

Output at 4:5, 2K resolution.

3 Things I Learned the Hard Way

Name what you want preserved, not just what you want changed. When I’m doing a surgical edit, “keep everything else identical” is not specific enough. Name the elements: background, layout, typography, colour of every other button. The more explicit you are about what stays, the more precise the edit.

Tell it where the output is going. “For a Notion page,” “for a 16:9 slide,” “for Slack” - this changes how Nano Banana 2 handles margins, text sizing, and resolution. I skip this when I’m rushing and then I’m always resizing manually afterward.

512px is your cost control lever. For high-volume generation, generate at 512px and upscale separately. You get the Nano Banana 2 reasoning quality at Nano Banana 1 pricing. The difference between 512px and 1K is invisible at screen resolution in most presentations.

That’s the full Nano Banana 2 playbook. Save the prompts, try the Frankenstein mockup first, and let me know what you end up building.

I watched Dylan Patel’s latest interview on the future of AI and chips. Four things stayed with me:

1. Agent orchestration is rewriting white-collar work. Claude Code and tools like it are no longer just coding assistants. They’re becoming full orchestration systems that let non-programmers build complex software and data pipelines in natural language. Companies using this are operating 10x faster than before. The junior developer market is getting nuked in the process.

2. The US is losing the military AI race. The US military is running on older classified models. China is deploying the newest ones for autonomous drone swarms right now. And while Anthropic is pushing back on government demands that could enable mass surveillance, OpenAI has already moved in to secure Department of War contracts.

3. Hardware constraints are locking in cloud AI dominance. Global shortages in memory and 3nm/5nm silicon are already forcing major mobile companies to cut smartphone production. The industry is routing everything scarce toward data center clusters. Local AI inference is becoming more expensive and less practical, not less. Closed-source cloud wins by default.

4. Scaling laws are still intact, and the ASI race is accelerating. Pre-training and RL scaling are both holding. Costs keep dropping year over year, capabilities keep jumping. Anthropic is the consensus leader right now based on revenue and model quality. OpenAI remains the contrarian pick to reach ASI first.

Watch Full Conversation

That’s all for today. See you next week,

Aakash

PS: You are receiving this email because you were subscribed to Aakash’s Product Growth Newsletter, from where the first 16 issues of the AI Update were sent. I have now given it its own home.

Share

Every single person I’ve turned on to Claude Code and Claude Cowork has been happy I did. I really think it’s not an exaggeration to say:

They’re the most important new professional software created in years.

But the issue everyone runs into is starting with them. Today, I’m open-sourcing everything you need to start with them well. So you have no excuse!

That’s today’s deep dive.

But first, there’s been an absolute torrent of interesting AI news. I’ve crossed 200M impressions on X alone the past month covering it. And today, I’m bringing you the web’s highest signal summary of the last week. But first…

Nimbalyst: The Visual Workspace for Claude Code and Codex

If you’re using Claude Code in the terminal, you’re missing half the picture.

Nimbalyst gives you visual editors for everything your agent touches. Markdown, CSV, mockups, diagrams, data models, code. You see exactly what the agent reads and edits, review changes as diffs, and approve before anything lands. No more switching between terminal, VS Code, and Figma:

Try Nimbalyst

There’s a million AI news articles, resources, tools, and fundraises every week. You can’t keep track of everything. Here’s what mattered - the 1 big story and key news.

Nano Banana 2 Changes the Math on AI Image Generation

Remember when Nano Banana went viral last August? Then Nano Banana Pro dropped in November with studio-quality output. Now Google shipped Nano Banana 2. And it combines the quality of Pro with the speed of Flash:

That makes it the biggest drop of the last week. The real story is three things that actually matter for builders:

1. Text in images finally works

If you’ve ever tried putting text inside an AI-generated image, you know how painful it was. Blurry, misspelled, wrong fonts. Every time. Nano Banana 2 can generate posters, banners, and infographics where the text is actually readable. It even handles translation across languages without breaking the layout. If you’re building anything customer-facing with AI visuals, this was the #1 blocker. It’s gone now.

2. The iteration speed is completely different

You can now test a bunch of variations quickly at low resolution, pick the one that works, and scale it up to 4K. The feedback loop that used to take minutes now takes seconds. For anyone doing product mockups, marketing assets, or even just exploring ideas visually, this changes how you work.

3. The API cost dropped 4x

It also pulls real-time web knowledge from Gemini during generation. So it can reference current events and real brand elements instead of hallucinating something from 2023. Surprisingly useful for marketing use cases.

If you read my Nano Banana Pro prompts, those all still work. Just better and faster now. The API is live. Go test it.

Key News

• Anthropic told the Pentagon no. The Department of Defense wanted Claude’s safety limits removed. Anthropic walked away from a $200M contract. And it got labelled a “Supply Chain Risk” as a result. That hurts their enterprise deals, but hasn’t stopped them going from $14B to $19B ARR in about a month. It is one of the most effective consumer brand strategies in AI.

• Perplexity is now integrated into all Samsung Galaxy S26 phones. Their Sonar API powers Bixby’s search. Samsung’s own assistant now runs on Perplexity’s infrastructure.

• DeepSeek’s upcoming V4 model was trained on smuggled Blackwell chips. The DeepSeek narrative was always “2,000 H800 chips, $5.6 million, and Chinese ingenuity beat $100 billion in American spending.” That story just collapsed.

Tool Updates

• Perplexity launched Computer. Think of it as “OpenClaw for the less technical.” It orchestrates 19 AI models to complete entire workflows. You describe what you want done, it breaks it into subtasks, picks the best model for each one (Claude for reasoning, Gemini for research, Nano Banana for images, Grok for speed), and runs in the background.

• Cowork got Scheduled Tasks. Claude stopped being software you talk to and became software that works while you sleep.

Fundraising

• WorkOS raised $100M at a $2B valuation. They sell SSO and directory sync. Sounds boring until you see the customer list: OpenAI, Anthropic, xAI, Cursor, Perplexity, Replit, Vercel. Every company building next-gen enterprise software runs on WorkOS. That’s a bet on the entire AI application layer.

• OpenAI raised $110B. The real signal is in two technical terms most people scrolled right past. Microsoft gets “stateless API.” Amazon gets “stateful runtime environment.” OpenAI just handed the growth segment to Amazon while leaving Microsoft with the architecture that’s becoming the commodity layer.

I Open-Sourced Claude Code Setup and 5 Key Skills

I pay for Google AI for Nano Banana + NotebookLM. Gemini itself is still much worse than Claude.

That’s because Claude Cowork and Claude Code are absolute powerhouses for any knowledge worker.

But the problem everyone runs into is starting with Claude is starting. So I’m open sourcing a free ‘0 to 100’ guide with 2 new Github repos today.

Deep Dive Outline

I’ve put together the ultimate startup guide for Claude Code and Cowork:

1. Building Claude’s Memory

2. 5 Starter Skills

3. Claude Code Setup

4. Where to Go From Here

Subscribe now

1. Building Claude’s Memory

So let’s start with the new Import Memory feature:

You go to Settings > Capabilities, find the Memory section, and select “Start import.”

It has you put this prompt into your existing AI (Google or OpenAI likely):

I’m moving to another service and need to export my data. List every memory you have stored about me, as well as any context you’ve learned about me from past conversations. Output everything in a single code block so I can easily copy it. Format each entry as: [date saved, if available] - memory content. Make sure to cover all of the following — preserve my words verbatim where possible: Instructions I’ve given you about how to respond (tone, format, style, ‘always do X’, ‘never do Y’). Personal details: name, location, job, family, interests. Projects, goals, and recurring topics. Tools, languages, and frameworks I use. Preferences and corrections I’ve made to your behavior. Any other stored context not covered above. Do not summarize, group, or omit any entries. After the code block, confirm whether that is the complete set or if any remain.

Then you copy the results into Claude’s memory import feature (not a regular chat), and Claude will update its memory from your prior AI.

If you haven’t developed a memory with AI, open a chat with Claude and complete the following prompt:

I'm [name]

My most important goals are: [1. X, 2. Y]
I will be using Claude to [X, Y, X] 
My profession is [X]
My biggest roadblock at work is [X]
One thing I want to get better at is [X]

Add this information to your memory about me so you can be my personalized AI copilot that 10x's my ability to reach my goals with the goal of model obsolescence and self-sufficiency.

This might seem like a small thing, but it makes a huge difference to the quality of the responses you receive.

2. 5 Starter Skills

The most important feature that Claude has is skills. This is the big differentiator from ChatGPT and Google.

ChatGPT Projects and Gemini Gems are just wrappers on top of their AI. Skills are how you actually train an AI to do work for you.

But most people struggle with starting with skills. So steal mine:

Get the Claude Skills

Setup takes 30 seconds:

git clone https://github.com/aakashg/pm-claude-skills.git
cp -r pm-claude-skills/skills/* /path/to/your/project/.claude/skills/ 

Open Claude Code. The skills load automatically.

As you can see on the Github, there are five of them:

Here’s what makes skills different from just prompting Claude: consistency. A good skill encodes your exact workflow - decision points, output format, anti-patterns to avoid. It produces the same quality output on attempt 50 as attempt 1.

Skills ideally would be automatically invoked, but aren’t always, so when you want to use them say something like this in your prompt:

Use the LinkedIn Post Writer Skill

The repo also includes a SKILL-TEMPLATE.md so you can build your own.

I wrote a deep dive breaking down 6 additional skills with the reasoning behind each one and how I built them:

Read the full breakdown here

3. Claude Code Setup

This is the one people have been asking about.

But you can’t accomplish real PM automation with a vanilla Claude Code setup. You need a CLAUDE.md file that gives Claude persistent context about who you are, how you work, and what you expect.

I iterated on mine over 100 times. Today I’m open sourcing the starter version: pm-claude-code-setup

Get my Claude Setup Github

What’s inside:

pm-claude-code-setup/
├── CLAUDE.md                           # Your persistent context file
└── .claude/
    └── skills/
        └── prd-writer/
            └── SKILL.md                # Your first skill

The CLAUDE.md file tells Claude:

• Your role, company, and product (so it stops asking)

• Your north star metrics (so recommendations align with what matters)

• Your PRD structure (so every PRD follows the same format)

• Your writing style rules (so output sounds like you, not a corporate AI)

• Sub-agent roles — say “review as engineer” or “review as skeptic” and Claude shifts perspective

The PRD Writer skill:

• Asks 3-5 clarifying questions before generating anything (this alone saves hours of back-and-forth)

• Follows a strict structure: hypothesis, problem with data, strategic fit, solution, success metrics, non-goals

• Flags gaps with [NEED: data from X] placeholders instead of making things up

Setup in 60 seconds:

1. Clone the repo

2. Copy CLAUDE.md to your project root

3. Copy the .claude/ folder to your project

4. Fill in the [FILL IN] sections with your info

5. Open Claude Code. It loads automatically.

That’s the difference between Claude guessing what you want and Claude knowing what you want.

I built a PM OS in Claude Code going much further down this path if you want to really upgrade your setup:

Get the PM OS

4. Where to Go From Here

Start using Claude Cowork and Claude Code! They are incredibly powerful.

For Cowork:

• Connect your company data

• Use it to help you write something

For Claude Code:

• Build an automated tool you need

• Improve your Claude.MD file over time

You now are equipped to use some of the most powerful software ever created.

You are receiving this email because you were subscribed to Aakash’s Product Growth Newsletter, from where the first 15 issues of the AI Update were sent. I have now given it its own home.

Share

Of all the AI tools I use that I see the least other people using, NotebookLM tops the list. It’s the entire reason I pay for Google AI.

It is not getting the attention it deserves. And they just launched their best feature yet: prompt based slide revisions. So, I had to make you a complete guide.

But first, there’s a million pieces of AI news every week. You can’t keep track of everything. So I’ve done it for you. Here’s what actually mattered.

In partnership with Polymarket

The Top News of the Week: Gemini 3.1 Pro

Google just released Gemini 3.1 Pro.

But despite that release, Polymarket doesn’t think it will have the best AI model at the end of the month:

What’s going on? That’s because in the Arena AI rankings, Opus-4.6 is still winning the text category (no not the Sonnet 4.6 model they released this week).

That’s the story of Gemini 3.1 Pro. It’s not the best text model (it is #3, higher than GPT-5.2). But it is a heckuva an image model. There, Gemini 3 and 3.1 still hold the #1 and #2 rankings.

This model can generate website-ready, animated SVGs from a simple text prompt. Since these are built in pure code and not pixels, they stay crisp at any scale with incredibly small file sizes:

It can also translate the literary style of a novel into the design of a personal portfolio site. And it can create interactive 3D simulations with real-time hand tracking and generative audio.

It’s worth noting that the context window stays at 1M tokens and the output limit jumped to 65K tokens. That means you can feed it a medium-sized code repository and get back a 100-page technical manual in a single turn.

Google is iterating fast. This is the first time they’ve done a 0.1 increment instead of waiting for a 0.5 release.

I’m waiting for May/June when they have that next big release:

The Other Top 5 Stories of the Week

1. Anthropic shipped Claude Sonnet 4.6 with a 1M token context window and what I think is the biggest impact is the huge impact this has on SMB automation. Frontier performance just got cheap. But I don’t recommend it as a model for personal LLM use. Pay up for Opus.

2. Anthropic launched Claude Code Security, and it just impacted the $15B market price in a major way. Everyone in AppSec is no longer selling “we find more bugs.” They’re competing against an AI that finds them and writes the patches in the same session.

3. Figma now imports Claude Code UI as editable design layers. This closes the last excuse PMs have for not shipping polished UI from AI code.

4. OpenClaw creator Peter Steinberger joined OpenAI and it says a lot about building in Europe. He said it himself: “In the USA, most people are enthusiastic.
   In Europe, I get insulted, people scream REGULATION and RESPONSIBILITY.”

New Tools Worth Your Time

It’s not just NotebookLM that makes a Google AI subscription worth it:

• Google Labs Pomelli generates studio-quality product photography for free. They just bundled what others sold as a paid product for free with their AI subscription.

• Google Lyria 3 generates 30-second songs with vocals, instruments, and cover art inside the Gemini app. Watch out for that Suno valuation.

Notable Fundraising

• World Labs raised $1 billion from Nvidia, AMD, a16z, and Fidelity. I wrote about Fei-Fei’s story, crossing 1M views and getting response from her.

And now on to today’s deep dive:

Complete Guide to NotebookLM

NotebookLM started as a tool to search your notes. Now it generates podcasts, videos, slide decks, infographics, flashcards, quizzes, mind maps, and reports.

48 million monthly visits. 120% quarter-over-quarter user growth in Q4 2024.

Most people think of it as a PDF summarizer. It’s a research-to-product pipeline if you use it right.

I used one notebook to go from zero knowledge on a topic to a working product prototype, customer discovery materials, and a product deck in a single afternoon.

Let me show you how.

Today’s Post

Here’s your complete guide to NotebookLM:

1. Why NotebookLM Is Architecturally Different From Every Other AI Tool

2. The 3 Killer Features (And How to Actually Use Them)

3. From Research to Product Deck in One Afternoon

4. Power Use: Prototypes, Gems, and Automation

5. The AI Stack That Actually Works

Subscribe now

1. Why NotebookLM Is Architecturally Different From Every Other AI Tool

Every other AI tool you use answers from the entire internet plus its training data. You upload a document, ask a question, and the model blends what you gave it with everything it already “knows.” The output is a smoothie. You can’t tell which insights came from your data and which the model made up.

NotebookLM answers only from your sources. Nothing else.

When it doesn’t have the answer in your sources, it says so. It doesn’t guess. It doesn’t fill gaps with plausible-sounding training data. That one architectural decision is why I pay $20/mo for Google AI.

What Just Shipped

Google launched prompt-based slide revisions. Before, NotebookLM generated take-it-or-leave-it decks. Now you iterate through conversation. “Make slide 3 more visual.” “Add a comparison table to slide 7.” Slide generation without revision was a toy. Slide generation with revision is a workflow. That feature is why I wrote this guide.

They also shipped video overviews, branded infographic generation with your own assets, data table exports to Google Sheets, and a mobile app for audio overviews on the go.

2. The 3 Killer Features (And How to Actually Use Them)

NotebookLM has 10+ output formats. Three of them are genuinely differentiated from anything else on the market:

1. Slide Decks

2. Sole-Sourced Answers

3. Infographics With Brand Assets

Let me walk through each.

Killer Feature #1: Slide Decks With Prompt-Based Revision

Here’s what surprised me: you can click “Slide Deck” in the Studio panel with zero prompting and get a genuinely good deck. NotebookLM reads your sources, figures out the narrative arc, picks the key data points, and generates 12-15 slides with speaker notes. The visual quality is strong thanks to Google’s Imagen model. The storytelling structure gives me ideas on how to present messy information I wouldn’t have thought of myself.

That alone puts it ahead of every other AI slide tool I’ve used.

But now with the new prompt-based revision, you can make them actually presentation-ready. Generate the deck, then use the “Revise” button to iterate. “Rewrite the speaker notes for a more technical audience.” “Swap slide 4 to a comparison layout.” “Make the conclusion more action-oriented.” Same revision loop you’d have with a human collaborator.

And if you do want to prompt from the start, here’s the power-user workflow: create a note called “slide deck specification” with your audience, slide count, structure, visual style, color scheme. This becomes a reusable playbook you reference every time. Copy the spec, paste it with your content focus, and generate.

When your VP asks “where did this number come from?” you can answer, because every claim traces back to your sources.

Killer Feature #2: Sole-Sourced Answers With Citation Tracing

I covered the architecture in Section 1. Here’s how to actually use it.

Selective source toggling. Toggle individual sources on and off before asking. Compare just the academic research. Then just the practitioner guides. Then both. You turn a flat pile of documents into a structured, interrogatable knowledge base.

The gap analysis technique. Ask: “What are the five most important questions a [role] should be asking about [topic] this year?” Forces prioritization over summarization. Then: “What are the missing sources to dive deeper?” Identifies blind spots in your own research.

The note-to-source loop. Save any answer as a note, convert it to a source, then generate outputs from that curated answer. That loop is how you go from raw research to polished content in layers.

Killer Feature #3: Infographics With Brand Assets

Upload a brand character or visual element as a source. Tell NotebookLM to use it throughout the infographic. It weaves your asset into the design.

The workflow: upload brand character → optionally generate infographic copy first as a note → convert note to source → select brand character + copy source → click Infographic → pick orientation, style (hand-drawing, whiteboard marker, grid layout), and specify the character in your prompt.

Here’s a recent infographic it generated for the podcast:

It’s pretty good, right? Easily the best in AI right now.

3. From Research to Product Deck in One Afternoon

Here’s why NotebookLM is different from the 50 other AI tools you’ve tried this year: it’s an ecosystem. One notebook holds your sources, generates your mind map, creates your podcast, builds your slides, exports your data tables, and feeds into Gemini for prototyping. Everything talks to everything else. You never start over.

Let me show you how that plays out on a real project.

I’ve been watching the Generative Engine Optimization (GEO) space. The tools that exist today mostly track visibility but don’t tell you what to fix. There’s a gap. I wanted to build a quick prototype and see if people cared.

The workflow breaks into four phases: Research → Learn → Analyze → Ship.

Phase 1: Research (4 minutes)

Deep Research in the Studio panel. Typed my question. Four minutes later: a report covering 55 resources with citations, imported directly as a source.

Then a mind map to see structure before reading anything. The map showed where tools clustered (monitoring visibility) and where gaps existed (nothing for telling you what to fix, connected to revenue). Downloaded the PNG. Used it in every GEO conversation that week.

Phase 2: Learn (while making dinner)

I had 55 sources. I wasn’t going to read them.

Generated two audio overviews. A Deep Dive while cooking, where the hosts covered how AI search differs from Google, the three strategy buckets, and why most brands are getting it wrong.

Then a Debate to stress-test: is GEO a new problem or just SEO with extra steps? Hearing both sides argued out loud forced me to form my own position.

Then Interactive Mode. Clicked “Join” and asked: “How does this apply to B2B SaaS specifically?” The hosts pivoted the entire discussion, pulled from practitioner guides, gave specific examples. That’s not a podcast. That’s a conversation with your research.

Phase 3: Analyze (finding the product opportunity)

Selective source toggling. Turned on just the tool docs to compare what exists. Then just the academic research. Then both.

Three questions.

What’s the biggest gap? Tools track mentions but none close the loop: track → diagnose → tell you what to fix → measure revenue impact.

What exists already? Some content auditing, nothing connecting audit to action.

What would the MVP look like? NotebookLM gave me a cited answer: prompt-level tracking, gap diagnosis, prioritized fix list, before/after scoring. That answer became the product spec.

Phase 4: Ship (customer discovery + product deck)

Generated an infographic from the Studio panel. Used it in conversations with 5 SaaS founders. Instead of 10 minutes explaining GEO, I showed the infographic and asked: “Is this happening to you?”

Generated a Whiteboard-style Video Overview: “why brands are invisible in AI search.” 4-minute narrated video. 3 minutes to generate.

Slide Deck from the Studio panel with my specification note. Data Table to compare six GEO tools across features, pricing, engines tracked, and gaps. Exported to Sheets. Screenshot into the deck.

A real product deck from real research in about an hour. Research → Learn → Analyze → Ship. Every step fed the next. Every output in the same notebook. I never lost context.

4. How to Connect NotebookLM to Gemini (Prototypes, Gems, and Automation)

Everything so far lives inside NotebookLM. Here’s where you break out.

Connecting Them

Open Gemini → click “+” → select “NotebookLM” → pick your notebook → click “Add.” Gemini now has read-access to everything. Every answer includes citations.

And unlike NotebookLM, Gemini can search the live web while referencing your notebook, so you can ask what you’re missing and add the new sources back.

Building a Working Prototype

With my GEO notebook attached, I asked Gemini to build a working GEO Audit Tool using Canvas. It built a live, interactive app. Type a brand name, enter test prompts, see audit results with recommendations.

I iterated for 40 minutes. Adjusted scoring weights, added explanations, cleaned up UI, all through conversation. Gemini is genuinely good at front-end work. I covered AI prototyping workflows for paid subscribers.

The mental model: NotebookLM is your second brain (facts, sources, truth). Gemini is your builder (interactive body).

Gems: Persistent AI Advisors

Create a Gemini Gem with your NotebookLM notebook as a knowledge source. It auto-syncs when you add new sources. It scales past limits, Gems cap at 10 files but through NotebookLM access hundreds. And instructions persist across sessions, come back weeks later, pick up where you left off.

I now have a “GEO Advisor” Gem I use whenever someone asks about AI visibility. New research goes into the notebook. The Gem learns automatically.

Automating at Scale (MCP Server)

NotebookLM doesn’t have an official API yet, but the community built open-source MCP servers connecting Google Antigravity (or Claude Code, or Cursor) to NotebookLM.

uv tool install notebooklm-mcp-cli
nlm setup add antigravity

One prompt creates a notebook, adds sources, runs deep research, and generates outputs. I’ve started building pre-populated audit notebooks for different verticals. Each one takes a single prompt. This is how you scale a research process into a service.

5. The AI Stack That Actually Works

I get asked “should I use NotebookLM or ChatGPT?” twice a week. Wrong question. Here’s the stack I actually use and why each tool earned its spot:

• NotebookLM owns three things: sole-sourced research, slide decks, and infographics. The only tool I trust when “where did you get that?” matters. Client deliverables, strategy presentations, competitive analysis.

• Claude Code is my daily driver for deep work. I built my entire PM Operating System inside Claude Projects.

• v0 or MagicPatterns for quick UI prototyping, when you need a visual in 30 seconds.

• Speechify for dictation. 3x typing speed, AI cleans up grammar. If you’re not using a dictation tool in 2026, you’re leaving speed on the table.

The flow: Speechify gets raw thoughts down → NotebookLM does deep research → NotebookLM synthesizes and generates outputs → Claude Code handles deep writing → v0 mocks quick UIs.

My full AI toolkit breakdown goes deeper on each.

Final Words - Google’s Disconnected AI Strategy

Google has Gemini the model, Gemini the chatbot, Gemini the API, Google AI Studio, Vertex AI, Antigravity, and NotebookLM. Each does something the others can’t. The connections feel stitched together rather than designed.

The NotebookLM-to-Gemini handoff from Section 4? Powerful. But you have to know it exists. There’s no onboarding that says “hey, you just built a research notebook, want to turn it into an app?” Compare that to Anthropic, where Claude Projects, Claude Code, and Cowork live under one roof.

NotebookLM is still the one Google product I’d genuinely miss if it disappeared. The sole-sourcing is architecturally unique. The slides just got actually good. The audio overviews remain the best way to internalize research without reading.

Google has the tools. They need the map.

YC interviewed Anthropic’s Boris Cherny, and it was a great podcast.

1. The “software engineer” title is fading

Since Claude Code’s internal rollout, productivity per engineer at Anthropic is up 150%. Cherny says 90-100% of code is now written by the model. Designers and PMs are coding. The job is shifting from writing code to writing specs and talking to users. That’s not a prediction. It’s already happening inside one of the top AI labs.

2. Build for the model six months from now

Anthropic doesn’t build around current model limitations because those limitations tend to disappear with the next release. All that “scaffolding” code developers write to patch model errors? It becomes technical debt the moment a better model ships. Cherny’s advice is to leave those gaps open and let the next model close them.

3. Watch what users do, not what your roadmap says

Claude Code wasn’t planned. Engineers were manually telling Claude to think before coding, so Anthropic built “plan mode.” Finance and sales teams started using the terminal app for non-engineering work, so they built the Cowork GUI. The product grew from watching real behavior, not from a strategy doc. You can’t force people to do something new. You make what they’re already trying to do easier.

Watch Full Conversation

That’s all for today. See you next week,

Aakash

P.S. You can pick and choose to only receive the AI update, or only receive Product Growth emails, or podcast emails here.

Claude Cowork is not getting the hype it deserves. This is an LLM harness that lets your AI do all the work for you in systems you control:

• Working in Excel or Powerpoint

• Building Google Docs or Notion pages

More people should be using it. Today I’m giving you the ultimate guide to Claude Cowork.

But before we get there, let’s get you caught up on the week’s AI news that matters.

Lev8: Your AI-powered Connection Engine

Stop guessing who to talk to, and start connecting with the right people. I used Lev8 to easily find the people I want, and their contact information:

Just tell Lev8 who you’re looking for, and it’s like having a personal assistant scouring the web in real-time, bringing back only the most relevant, verified contacts

I’d love for you to try out Lev8. Here’s an exclusive invite just for you: HIIMLEV8

Get Early Access

There’s a million AI news articles, resources, tools, and fundraises every week. You can’t keep track of everything. Here’s what mattered.

Top News

Seedance 2.0 is creating major buzz even before its US release

ByteDance released Seedance 2.0 this week. And it used the Sora playbook: going viral with massive, copyrighted content drops.

Irish filmmaker Ruairi Robinson typed a two-line prompt and got Tom Cruise fighting Brad Pitt on a rooftop. Within hours of restricted release, people were generating alternate endings to Game of Thrones, Optimus Prime fighting Godzilla, Spider-Man scenes with just a single prompt.

The model isn’t publicly available to the world yet. It’s accessible through ByteDance’s Jimeng (Dreamina) platform in China, with a global launch expected within the next two weeks. And it’s already the most talked-about AI model of the month from demo clips alone.

On benchmarks, Seedance 2.0 leads in motion consistency and temporal coherence across most video generation tasks. In head-to-head comparisons against Sora 2, Kling 3.0, and Veo 3, it wins on controllability and resolution. Sora 2 still edges it on pure physics simulation. Kling handles fast action sequences better. But for cinematic storytelling and brand creative, Seedance 2.0 is the new bar.

API access through partners like FLUX Context is confirmed for February 24th. Global public launch is expected late February or early March.

This is the moment AI video crosses from “impressive demo” to “usable production tool.” If you’re building anything that touches content creation, product demos, or marketing automation, pay attention.

Other Key News

• Three major Chinese LLMs released: Qwen-3.5, GLM-5, and Minimax-M2.5. The Chinese open-source machine keeps churning out models almost as good as Anthropic and OpenAI at a fraction of the cost.

• OpenAI upgraded Deep Research to GPT-5.2. Site-specific search, app integrations, full-screen reports. Each feature alone is incremental. Stacked together, they turn a chatbot into a research workflow that’s hard to leave. The model upgrade gets the headline. The workflow lock-in builds the moat.

• AI safety at Anthropic might be in peril. The head of Safeguards Research resigned, saying pressures to ship are winning over values. Three other safety researchers left in the same month.

Resources

• Andrej Karpathy shared a complete GPT in 243 lines of Python. Training loop, inference, optimizer, attention. The entire conceptual engine powering hundred-million-dollar training runs fits in fewer lines than a terms-of-service page.

• AI doesn’t reduce work, it intensifies it. HBR piece worth reading if you manage a team adopting AI tools.

Tools

• KimiClaw - Kimi released a cheap version of OpenClaw.

• Happy Capy An agent-native computer in your browser, where you can describe what you need and it handles the rest. No setup, no learning curve. #1 on Product Hunt this week

Fundraising

• Anthropic raised $30B at a $380B valuation. Led by GIC and Coatue. Second-largest private tech raise ever. Claude Code alone has a run-rate above $2.5B.

• Simile AI raised $100M led by Index Ventures to build AI simulations of human behavior. Backed by Fei-Fei Li and Andrej Karpathy.

• Apptronik raised $520M to boost production of its humanoid robot Apollo.

Events

• I’m teaching a live webinar on designing agents that build agent teams. If you’re a PM or builder working with AI agents, this is for you.

Enroll here for free

And now on to today’s deep dive:

The Web’s Deepest Guide to Claude Cowork

People are still not using Claude Cowork to its full potential.

I get it. When I first covered Cowork a month ago, it was brand new. Mac only. Cool but limited. A lot of people tried it, organized some files, and moved on.

That would be a mistake.

In the last four weeks, Anthropic shipped plugins, Windows support, scheduled tasks, browser automation, and direct integrations with Excel and PowerPoint.

Cowork went from “neat desktop assistant” to what I’d call your first real AI employee. It reads your files, browses the web on your behalf, logs into your accounts, builds spreadsheets with working formulas, creates decks that match your company template, and runs tasks on a schedule while you sleep.

If you’re not using this, you’re doing work that a machine can already do better and faster.

I’m not going to walk you through the basics again. If you need that, go read my earlier deep dive. What I want to do today is give away my exact workflows for the features that actually replaced hours of my week. The stuff nobody is talking about yet.

Here’s what you need to know:

Continue Reading Online

1. Claude in Excel

This is the one that hit me the hardest.

I had a folder full of receipt screenshots. Uber rides, hotel bookings, software subscriptions, client dinners. The kind of stuff you throw into a folder and tell yourself you’ll organize “later.” I used to hand all of this off to a financial assistant.

Now I open Cowork and run this prompt:

Open all pdf in /invoices. Extract tables (date, vendor, amount, category). Create "Transactions_2026.xlsx" with clean columns, normalized dates, and confidence scores per row. Add a pivot sheet summarizing monthly spend by category and a chart on sheet 'SummaryChart'. Save workbook to Cowork folder.

Now I just keep dropping new receipts into that folder and run the same prompt again. No more waiting on anyone.

But here’s where it gets wild. Claude isn’t just available through Cowork anymore. It now lives inside Excel as a native add-in.

How to set it up:

1. Open Excel

2. Go to Insert, then Get Add-ins (on Windows) or Tools, then Add-ins (on Mac)

3. Search “Claude by Anthropic”

4. Click “Get it now” and sign in with your Claude account

5. The Claude sidebar shows up. You’re live.

Quick shortcut to open it anytime: Ctrl+Option+C on Mac, Ctrl+Alt+C on Windows.

Once it’s running, Claude reads your entire workbook. Every tab, every formula, every cell dependency. When it explains something, it gives you cell-level citations so you can actually verify the logic. It traces formulas back to their source inputs. It understands relationships across tabs.

Here’s a real example. I had a marketing budgeting sheet that I was building out for quarterly planning. The raw data was there across tabs, spend by channel, anticipated revenue targets, monthly actuals. But I needed the kind of metrics that actually tell you whether your spend is scaling efficiently or completely ballooning.

Here’s what I asked Claude to do:

Look at the Budgeting sheet. Add these computed metrics for each month:

- Marketing Spend as % of Sales (total marketing budget / anticipated sales for that month)
- CAC by channel (spend per channel / new customers acquired from that channel)
- ROAS by channel (revenue attributed / spend)
- Month-over-month delta for each metric

Then create a performance graph showing all key metrics trended month over month. Pick the best chart type for each metric. Use conditional formatting on the summary row, green for improving and red for declining.

Every formula was clean. Every chart was the right type for the data it was showing. I reviewed it, made one small label tweak, and it was done.

That’s the real ROI here. It’s not just that Claude writes formulas faster than me. It’s that it makes the visualization decisions I used to burn time thinking about. The kind of decisions where you stare at the screen for 15 minutes going back and forth between chart types before picking one. Claude just picks the right one and moves on.

I think this makes a lot of the repetitive spreadsheet grunt work that junior analysts and operations teams spend hours on every week completely unnecessary.

2. Claude in PowerPoint

The setup is the same as Excel. Search “Claude by Anthropic” in the PowerPoint add-in store. Sign in. Done.

Now, I’ve tried a lot of AI slide tools. Every single one creates slides that look like they came from a completely different company. Wrong fonts. Wrong colors. Generic layouts that scream “AI made this.”

Claude in PowerPoint is different because it reads your slide master before it does anything. Your layouts, your fonts, your color scheme. When it generates or edits slides, they actually match your template. That’s a small detail that changes everything when you’re building client-facing or executive decks.

The workflow that saved me a full afternoon:

I had a quarterly product review deck that my team had put together. It was 24 slides. The data was solid, but the storytelling was all over the place. Metrics were buried on slide 18 that should have been on slide 3. There was no clear narrative arc. Some slides had six bullet points where they needed two. The kind of deck that an executive looks at for 10 seconds and says “what’s the takeaway?”

Instead of spending three hours rearranging and rewriting, I told Claude:

This deck is a Q4 product review. Restructure it to be executive-ready. The story should follow this arc: 1) What we set out to do this quarter (goals), 2) What actually happened (results with key metrics upfront), 3) Where we fell short and why, 4) What we're doing next quarter. Move the most important metrics to the first three slides. Cut any slide that doesn't directly support the narrative. Consolidate bullet-heavy slides into cleaner formats. Keep the existing template and formatting.

Claude restructured the entire deck. It moved the revenue impact and adoption metrics to slide 2 where they belong. It consolidated three slides of bullet points about feature launches into one clean slide with a visual timeline. It cut four slides that were basically restating what earlier slides already covered. And it rewrote the headers on every slide so each one tells you the takeaway before you even read the details.

The final deck was 16 slides instead of 24. Tighter. Clearer. Every slide earned its spot.

That’s the real use case here. Most PMs don’t need AI to create a deck from scratch. They need AI to take their messy first draft and make it executive-ready. To restructure the story. To know that leadership cares about outcomes on slide 2, not methodology on slide 2. To cut the slides that aren’t pulling their weight.

The other workflow I keep coming back to is the Excel-to-PowerPoint pipeline:

1. Build the analysis in Excel using Claude (metrics, pivots, charts)

2. Open PowerPoint with your company template loaded

3. Tell Claude what to do with a prompt.

Raw data to polished deck. One prompt chain. No copying between applications, no reformatting, no alignment hell. Every consulting firm and services company charging you for this kind of analysis should be watching closely.

3. Claude in Chrome

This is the one that killed three subscriptions for me.

I was paying for Perplexity. I’d tested Manus and Comet and a handful of other browser agents. Claude in Chrome does what all of them do, but with one advantage that makes everything else feel incomplete: it uses your actual browser session.

Think about what that means. Manus runs everything in incognito mode. It can’t access your Gmail, your Salesforce instance, your internal dashboards, your Jira boards. It only sees what a logged-out stranger would see.

Claude in Chrome uses your real cookies, your real logins, everything you’re already signed into. It sees what you see. It can take actions on the pages you actually use for work.

How to install:

1. Go to the Chrome Web Store

2. Click “Add to Chrome”

3. Sign in with your Claude account

4. Pin the extension (click the puzzle piece icon, then the pin next to Claude)

It opens as a side panel inside your browser. From there, Claude can read pages, click buttons, fill forms, navigate between tabs, and manage multiple tabs at once.

Real example, competitive research:

I was prepping a competitive landscape slide for a product strategy review. Needed pricing, feature comparisons, and positioning across four competitors. Normally I’d open a dozen tabs, screenshot pricing pages, and manually build a comparison table in a Google Sheet.

Instead:

Visit [competitor1.com/pricing], [competitor2.com/pricing], [competitor3.com/pricing], and [competitor4.com/pricing]. Extract plan names, prices, and key feature differences. Organize everything into a comparison table.

Claude navigated to each site, pulled the data, and structured it into a clean table. What would have been 30 minutes of tab-switching and copy-pasting took about 3 minutes. I dropped that table into my strategy deck and moved on.

Where it gets crazy, paired with Cowork:

Chrome handles the web research. Cowork handles local file creation. Together they chain automatically.

I ran a full competitive analysis for a product planning session last week. Chrome visited five company websites and their G2 review pages. Cowork took all of that data and produced a 12-slide deck with positioning maps, feature comparisons, and pricing tables. I made a few tweaks and presented it the next morning. That entire workflow used to take me a full day.

Available on all paid plans. Pro users get Haiku 4.5 as the model. Max, Team, and Enterprise users can choose their model, including Opus 4.6.

4. Scheduled tasks and shortcuts

This is the feature that turned Cowork from a tool I use into a tool that works for me even when I’m not there.

Scheduled tasks let you set recurring browser workflows. Daily, weekly, monthly, whatever cadence you want. You set it once and Claude handles it from there.

Here’s what I have running right now:

Every Monday morning, Claude checks my top five competitors’ product changelog pages and blog posts for new launches or updates. It compiles a summary of what shipped, what’s new, and any pricing changes, then saves it to a markdown file in my research folder. Here’s the prompt:

Visit [competitor1.com/changelog], [competitor2.com/blog], [competitor3.com/changelog], [competitor4.com/releases], and [competitor5.com/blog]. For each, find any posts or entries from the last 7 days. Extract: product name, what shipped, pricing changes if any. Compile into a single markdown file organized by company. Save to /research/competitive-weekly-[date].md

By the time I sit down with coffee, the competitive briefing is already there. That used to be 45 minutes of manual browsing every single week.

I also have a weekly task that pulls usage metrics from our analytics dashboard, formats them into a tracking spreadsheet, and highlights any week-over-week changes that are above or below a threshold I set. It flags what needs attention so I’m not scanning rows looking for anomalies.

How to set one up:

1. Open Claude in Chrome

2. Click the clock icon in the upper right corner of the extension panel

3. Describe the task you want automated

4. Set the schedule (daily, weekly, monthly, or annual)

5. Claude runs it on schedule without you lifting a finger

Shortcuts are the other game changer. You record a workflow by doing it yourself once, step by step, and Claude learns to repeat it. Think of it like onboarding a new hire. You show them the process once, and then they can do it on their own every time after.

I recorded a shortcut for my weekly user feedback synthesis. Open our feedback tool, filter for the last 7 days, scan the top themes, pull the most common feature requests and complaints, and save a structured summary to my desktop. Used to take me 25 minutes of clicking and reading. Now I trigger the shortcut and Claude runs through the whole thing in about 90 seconds.

And the best part is..

You don’t even have to type to teach it. When you hit “Record workflow,” Claude watches your screen and listens to your microphone simultaneously. Click through the steps, narrate what you’re doing out loud and Claude learns both the actions and the context behind them. It’s the closest thing to onboarding an actual human teammate.

This matters because most PM work follows patterns. Same dashboards. Same data sources. Same formatting. Same stakeholder updates. Scheduled tasks and shortcuts mean you teach Claude once and then you never have to think about that task again.

Pre-approve the actions it can take before you start, and Claude will execute the full workflow independently within those boundaries. It only stops to ask if it hits something unexpected or a high-risk action.

5. Cowork plugins

Anthropic released something that I think most people completely overlooked: 11 open-source plugins for Cowork.

Each plugin is a complete package. Skills (the domain knowledge Claude draws on), slash commands (specific actions you trigger with /), MCP connectors (integrations with external tools), and sub-agents (specialized workers that run in parallel). All bundled together and configured for a specific business function.

How to install:

1. Open Claude Desktop and go to the Cowork tab

2. Click “Plugins” in the left sidebar

3. Browse the library and click any plugin to install it

4. Type / in any Cowork chat to see all available commands from your installed plugins

Here’s how I actually use it in practice. I have compiled all my youtube data in a folder and asked cowork to find gaps in it to improve my channel. I used the /explore-data command and claude found where my retention was dropping hard, what kind of thumbnail drove higher CTR and many more. Specific patterns I would have spent an afternoon finding in Claude found them in 90 seconds.

The real unlock is that every plugin is completely customizable. Don’t like how the Finance plugin structures a DCF model? Click “Customize” and tell Claude to adjust the methodology. Want to add your company’s specific feature categories or market segments to the Product plugin? Just describe what you need and Claude rewrites the plugin for you.

And they’re all open source on GitHub. Free to use. You only need a $20 paid Claude subscription.

This is what turns Claude from a general assistant into a specialized colleague who knows your domain, your processes, and your terminology. That’s a massive shift.

6. Parallel tasks

This feature doesn’t get talked about enough.

Cowork can now run multiple tasks at the same time across different tabs. One tab is doing competitive research while another tab is formatting your spreadsheet while a third tab is drafting a stakeholder update. All simultaneously.

I’m not exaggerating when I say this is the closest thing to having three junior analysts working at once.

Here’s how I actually use it. When I’m prepping for a quarterly planning session, I’ll open three Cowork tabs:

Tab 1: “Search the web for the latest market sizing data and analyst reports on [our product category]. Save a summary with sources to /research/market-update.md”

Tab 2: “Open the user feedback export in /data/ and categorize the top 20 feature requests by theme. Estimate effort level for each. Save as an Excel file with a pivot by theme.”

Tab 3: “Take the competitive pricing data in /research/ and create a positioning map showing where we sit versus competitors on price and feature completeness. Save as a PowerPoint slide using our template.”

All three run at the same time. I come back in 10 minutes and all three outputs are sitting in their respective folders. Ready for me to review and pull into my planning deck.

The reason this matters so much is that knowledge work is full of tasks that can happen in parallel but we do them one at a time because we only have one brain. We research, then we analyze, then we format, then we review. Always sequential.

Cowork removes that bottleneck. You describe the end state for three different workstreams and they all execute simultaneously. The time savings compound fast when you start thinking in parallel workflows instead of sequential steps.

You can also connect custom MCP connectors to specific tabs, so one tab is pulling data from your CRM while another is reading from Google Drive while a third is browsing the web. Each tab has its own context and its own tools.

That’s it for today’s deep dive. Finally, a podcast I found insightful.

I watched Nikita Bier’s latest interview and walked away with four insights I can’t stop thinking about:

1. There’s no silver bullet to growth. Bier is obsessed with not leaving a single stone unturned. He audits every step of the funnel, every button, every screen, every drop-off. He calls it “adversarial thinking,” trained from his early days as a hacker. If you’re looking for the one hack that’ll 10x your numbers, you’re thinking about it wrong. It’s a hundred small wins compounding.

2. X is operating like a 30-person startup. Despite being a global platform, X runs on about 30 product engineers with a flat org and minimal communication overhead. Bier describes working under Elon Musk as rebuilding the algorithm in two weeks and constructing a massive data center in two months. Whether you love or hate X’s direction, the velocity is undeniable.

3. Positive, niche social graphs go viral. Bier built TBH and Gas, essentially the same concept twice, by productizing anonymous compliments among teenagers. The key was density within specific school communities. He didn’t try to go broad. He went deep in one school, nailed it, then expanded. There’s a playbook in there for anyone building social products.

4. X wants to be a financial network. The long-term play isn’t social. It’s turning X into an everything app where you can bank and trade from your timeline. Features like Starter Packs (for finding niche communities) and improved link-engagement UI are building blocks toward that vision.

Watch Full Conversation

That’s all for today. See you next week,

Aakash

P.S. You can pick and choose to only receive the AI update, or only receive Product Growth emails, or podcast emails here.

Refer a friend

We had two major new model releases this week, Opus-4.6 and GPT-5.3-Codex. But the most important update, for my money was neither.

It was the Codex App.

And that’s why OpenAI hasn’t even released GPT-5.3-Codex in ChatGPT. You all are stuck on GPT-5.2, while Codex users are living in the future.

That’s today’s post.

AI Prototyping Like the Best

72% of the Fortune 500 is using Bolt as their AI prototyping tool of choice.

That’s because AI prototyping completely changes the game for product managers.

Instead of being stuck in the problem space forever, you can immediately get to work in the solution space.

And Bolt is the best tool for this, because you don’t just create throwaway prototypes. You create code that engineers can actually use because Bolt is connected to your code base and your design system.

Build with Bolt

There’s a million AI news articles, resources, tools, and fundraises every week. You can’t keep track of everything. Here’s what mattered.

Top News

Anthropic dropped Claude Opus-4.6. Everyone’s fixating on the 1M token context window, up from 200K. That’s roughly 1,500 pages of text, which means Claude can now read and reason across entire codebases without losing the plot halfway through.

But the context window isn’t the real story here. It is Agent Teams.

Multiple Claude instances now work in parallel on shared codebases. Instead of one agent grinding through tasks sequentially, a lead agent delegates to teammates who coordinate with each other while you sleep.

Anthropic says it’s a 4-5x throughput increase per developer. From what I’ve seen in Claude Code, I believe it.

Interestingly, they also released fast mode at 6x the cost and 2.5x faster output.

OpenAI’s response is equally fascinating in today’s deep dive.

Other Key News

• OpenAI launched Frontier platform. Enterprise hub to build, deploy, and manage AI agents across CRMs, databases, and internal tools.HP, Intuit, Oracle, State Farm, and Uber already onboard.

• OpenAI began testing ads. This is a stark departure from its competition.

• Anthropic ran Super Bowl ads mocking OpenAI’s decision to put ads in ChatGPT.

• Shopify CEO endorsed Pi over every funded AI coding tool.

• Nvidia paused gaming GPUs, data center revenue hit $51.2B last quarter while gaming was $4.3B. No new gaming GPU this year, first time in three decades.

• Claude in PowerPoint reads layouts, fonts, and matches corporate templates. Production-ready slides first try. No more fighting with alignment or brand guidelines.

Resources

• Anthropic revealed they built a 100K-line C compiler using 16 parallel Claude instances for $20K. The agents worked across 2,000 coding sessions, debugging their own code and coordinating without human intervention.

• Google’s PaperBanana: Closes the last gap in fully AI-generated papers.

Fundraising

• ElevenLabs raised $500M at $11B valuation, tripling their valuation in one year with Sequoia leading the Series D.

• Andreessen Horowitz raised $15B, their largest fund ever. $1.7B is going to AI infrastructure (Cursor, OpenAI, ElevenLabs, Black Forest Labs), another $1.7B for apps, $6.75B for growth.

• Nvidia nears $20B bet on OpenAI as the AI giant hunts a $100B+ round at $830B valuation. Jensen Huang confirmed this will be “the largest investment we’ve ever made.” Amazon is eyeing $50B, SoftBank discussing $30B.

And now on to today’s deep dive:

Codex Is the Best App OpenAI Has Ever Made

OpenAI counterpunched Opus-4.6’s launch by dropping GPT-5.3. But with catch: it’s Codex only.

Well, there’s a very good reason for that. Codex is the best app OpenAI has ever made. By far.

You tell it what to build, go to sleep and wake up to finished features waiting for review.

Before I show you how to actually use this, let me give you the full picture. I condensed everything into one cheat sheet below. Save it, print it out, then we’ll walk through the key pieces.

Today’s Guide

1. Getting Started with Codex in 5 Minutes

2. Powerful Workflows That Run While You Sleep

3. Built-In Automations (The Real Power)

4. Mid-Turn Steering (New in 5.3)

5. For Developers: Worktrees, Code Review Loops, and Cloud Execution

1. Setup Codex App in 5 mins

1A. Configure

Download the app from developers.openai.com/codex/app. Mac only for now ( Windows is coming soon)

Sign in using ChatGPT account or API key

Open a Git repo. This is critical. Codex works way better with Git. Worktrees require it. Clean diffs depend on it.

Choose GPT-5.3-Codex. GPT-5.2 is by default and weaker.

Test it. Create a new thread (⌘N). Type: “Summarize what’s in this project.” Hit enter.

1B. Setup your MCPs

This is the step most people skip. But it turns Codex from “coding tool” into a delegation engine for PMs.

MCPs are what let Codex talk to your tools. Without them, Codex can only work with files on your machine. With them, it can pull data from Zendesk, create tickets in Linear, check your Google Calendar, scrape competitor sites.

The question to ask yourself: what data sources do I use for my PRDs, strategy docs, and weekly updates today that I want Codex to have access to?

Here are the ones I recommend connecting:

• Support tickets (Zendesk, Intercom)

• Project management (Linear, Jira)

• Docs (Google Drive, Notion)

• Calendar and email (Google Calendar)

• Analytics (PostHog, Amplitude, Pendo)

• Code (GitHub MCP server)

You set these up in the MCP servers settings panel inside the Codex App:

For each tool, you’ll walk through an OAuth flow or paste in an API key. Codex will guide you through it. The whole process takes about 10 minutes for 3-4 tools.

Once your MCPs are connected, everything in section 2 becomes possible.

2. Workflows That Run While You Sleep

Workflow 1: Meeting Writeups → Action Items → Linear Tickets

As PMs, the power of the pen is one of the most important powers we wield. After a meeting, we send out the takeaways. We define what was decided.

Here’s how I use Codex for this entire loop.

Step 1: Record the meeting with Speechify, Otter, or whatever transcription tool you use. Drop the transcript file into your project folder.

Step 2: Create a new Codex thread (⌘N). Tell it to process the transcript:

Process this meeting transcript: [drag file in]

Output:
1. Summary (3-4 sentences — what was discussed and decided)
2. Key decisions made (with who made them)
3. Action items with owners and deadlines
4. Open questions that need follow-up
5. Draft Slack message for #product-updates

Tone: casual, direct, actionable. No fluff.

Hit enter. Codex will read the transcript and output formatted notes with action items extracted.

Step 3: If you have Linear MCP connected, take it one step further in the same thread:

Now create Linear tickets for each action item above.
Include: title, description, priority level, effort estimate.
Assign to the owner mentioned in the transcript.

Codex creates the tickets directly in Linear. No copy-pasting. No manual entry.

Why use Codex for this instead of an AI note-taking tool? Because Codex has your full project context. It knows your product, your team, your priorities. A generic tool gives you a generic summary. Codex gives you a PM summary with tickets that match your team’s format.

The whole process takes about 10 minutes. The old way took 1-2 hours between writing up notes, formatting them, and creating tickets manually.

Workflow 2: Competitive Intelligence

This used to eat 45 minutes every Monday. Pull up competitor sites, check changelogs, scan Product Hunt, read blog posts, look at G2 reviews.

Now I have Codex do it.

Step 1: Create a new thread. Give Codex the scope:

Analyze these competitors for changes in the past 7 days:
- [Competitor 1]: [website URL], [changelog URL]
- [Competitor 2]: [website URL], [changelog URL]  
- [Competitor 3]: [website URL], [changelog URL]

Also check their G2 pages for new reviews mentioning features we compete on.

Focus on: new features launched, pricing changes, positioning shifts, notable hiring.

For each change, tell me:
- What changed
- Why it matters to us
- Recommended action: ignore / monitor / respond

Output as a markdown report I can share with my team.

Step 2: Review the report Codex generates.

The real power here: turn this into a weekly automation (we’ll cover how in section 3). Schedule it for Sunday night. Wake up Monday with the report in your inbox.

Optional: If you have your analytics MCP connected, you can add a follow-up in the same thread:

Now cross-reference the features [Competitor X] launched against our own usage data.
- Do we already have similar functionality? What's our adoption rate?
- Are users requesting this in support tickets?

That’s competitive intel + internal data in one pass.

Workflow 3: User Research Synthesis

This is where Codex saves me the most time. The process is almost identical to what I do in Claude Code with my PM OS.

Step 1: Transcribe your interviews. Use Otter.ai, Zoom transcripts, or upload audio directly. Drop the transcript files into your project folder.

Step 2: Create a new thread. Tell Codex what to do:

Process these customer interview transcripts: [drag files in]

For each interview, extract:
- Key observations (what they said and did)
- Pain points (explicit problems mentioned)
- Workarounds (hacks they've built to solve the problem)
- Emotion signals (frustration, delight, confusion)
- Direct quotes worth saving (verbatim)

Then synthesize across ALL interviews:
- Group observations into themes (affinity mapping)
- Rank themes by frequency × severity
- Output:
  1. What to build (prioritized recommendations)
  2. What NOT to build (low-priority themes to ignore)
  3. Open questions (what needs more validation)
  4. Success metrics (how to measure if we solved it)
  5. Quote bank (15-20 quotes I can drop into PRDs)

Codex creates individual observation notes for each interview, then does the affinity mapping across all of them:

The output looks like what you’d get from a junior researcher who spent 6-8 hours on this. Except it takes 20 minutes.

Optional: Multi-Source Synthesis

This is where it gets really powerful. If you have Zendesk and PostHog MCPs connected, you can combine qualitative and quantitative data in the same thread:

Now cross-reference the top 3 pain points from the interviews against:
- Support tickets from Zendesk (past 90 days): how many users reported this?
- Product analytics from PostHog: what's the drop-off rate at each pain point?
- NPS responses: what are detractors saying about these issues?

Find patterns across all sources. Prioritize by signal strength.

This will output:

• Executive Summary (1 page for leadership)

• Full Synthesis Report (complete analysis)

• Quote Bank (for PRDs and presentations)

Workflow 4: PRD Creation + Multi-Perspective Review

As PMs, the PRD is still one of our most important artifacts.

Step 1: Create a new thread. Give Codex your feature context:

Create a PRD for [describe your feature].

Reference these documents for context: [drag in meeting notes, research synthesis, competitive analysis]

Include:
- Hypothesis (what we believe and why)
- Strategic fit (how this connects to company goals)
- User stories with acceptance criteria
- Non-goals (what we're explicitly NOT doing)
- Success metrics (how we'll measure impact)
- Rollout plan (phases, percentages, kill criteria)
- Edge cases and error states

The clearer you describe the feature, the better the output. Point to the meeting notes where you discussed it, the research synthesis from Workflow 3, the competitive analysis from Workflow 2. Codex uses all of it.

Step 2: Run multi-perspective reviews in the same thread:

Now review this PRD from 7 different perspectives:

1. Engineering: Is this technically feasible? What's the effort estimate? What are the dependencies?
2. Design: Does the user flow make sense? What's missing from the UX?
3. Executive: Does this align with company strategy? What's the ROI?
4. Legal: Any compliance, privacy, or regulatory concerns?
5. UX Researcher: What assumptions need validation before we build?
6. Skeptic: What could go wrong? What are we overlooking?
7. Customer Voice: Would a user actually want this? What would they say?

For each perspective, give me: top 3 concerns, questions to answer before building, and a confidence score (1-10) on whether we should proceed.

You get 7 perspectives on your work in 2 minutes. No meetings. No calendar Tetris. No waiting three days for feedback from legal.

After trying this process, you’ll never go back to the old way. PRD creation goes from 4-8 hours to about 90 minutes, and that includes the review cycle.

3. Built-In Automations (The Real Power)

Here’s the problem with every workflow above: they’re great when you remember to run them. But you’re busy. You forget. Or you do it Monday one week and skip it the next.

Codex solves this. It has automations built directly into the app. You schedule recurring tasks. They run automatically in isolated worktrees while you’re offline.

This is what turns Codex from “AI assistant I use sometimes” into “always-on teammate.”

How to Create an Automation

Open the Automations panel in Codex. Click “Create automation.” You’ll see:

• Name: Give it a clear name (e.g., “Weekly Competitive Intel”)

• Projects: Select which repo/project it runs against

• Prompt: Paste your prompt (same ones from section 2)

• Schedule: Pick frequency (daily, weekly) and specific days/times

Hit Create. That’s it. Codex runs this on schedule in the background.

The Automation I Recommend Starting With

Every day you do repetitive work that could be automated, but you don’t realize it’s repetitive until you’ve done it 10 times. You manually check metrics. You copy error logs into tickets. You update Slack when things ship. You scan PRs for TODOs that should become tickets.

These could all be skills. But you’re too busy doing the work to step back and notice the pattern.

So I set up this automation: every Sunday night, Codex reads how I used it that week and builds new skills based on what I kept asking for. My most common workflows become one-click commands.

Get Prompt here

Next week when I need to do a research synthesis, I just say “Use the Research Synthesis skill” instead of writing out the full prompt again.

Your Codex gets smarter every week based on how you actually use it. And this compounds. Every new thread reads the updated skills documentation. Every session starts smarter than the last one.

Other Automations Worth Setting Up

• Weekly competitive analysis: Run Sunday night, report ready Monday morning (use prompt from Workflow 2)

• Daily metrics summary: Pull key metrics from PostHog/Amplitude every morning, flag anything unusual

• Sprint retro prep: Every other Friday, pull completed Linear tickets and draft a retro summary

• Roadmap updates: Weekly check on Linear, update your Notion roadmap with what shipped

4. Mid-Turn Steering (New in 5.3)

This feature matters for PMs just as much as developers.

Before: if Codex was synthesizing your research in the wrong direction, you had to wait for it to finish. Then tell it to start over. Wasted 20 minutes.

Now: you see what Codex is doing in real-time. You realize it’s grouping feedback by user role when you wanted it grouped by pain point. You send a message while it’s working: “Stop. Regroup these by pain point, not by persona.”

Codex sees your message mid-task. Adjusts course immediately. Continues with your new instruction.

You just saved 20 minutes by catching the mistake early.

5. For Developers: Worktrees, Code Review Loops, and Cloud Execution

Everything above applies to developers too. But here’s what makes Codex specifically powerful for engineering workflows.

Cloud Execution

One of the most under-appreciated features of the Codex App is cloud execution. Codex can run tasks in isolated cloud containers instead of on your local machine. The agent still works on your repo but the execution happens remotely.

I close my laptop, come back after a coffee, and the refactor is done.

Enable Code Review in Codex Web to catch bugs before they ship. When enabled, Codex flags potential issues in your changes and suggests improvements, making automated code feedback part of your normal workflow.

Worktrees: Run 4 Agents Simultaneously

Worktrees are isolated copies of your Git repo sharing the same history. Four branches checked out simultaneously, each in its own folder, each running a different agent.

Setup: ⌘N → click “Worktree” instead of “Local” → pick your starting branch → Create. Repeat for each parallel task. One agent refactors auth, another fixes bugs, a third runs performance audits. No conflicts between them.

When a worktree agent finishes, you have two options:

Option 1: Tell Codex to create a branch, commit changes, push, and open a PR. It does everything automatically.

Option 2: Click “Hand off to Local” in the worktree thread. Review locally and merge when ready.

Archive completed threads to clean up old worktrees. Right-click thread → Archive.

Automated Code Review Loop

I submitted code with 3 problems last week: a type error, a style issue, and a broken test. Went to lunch. Came back 90 minutes later to find Codex had fixed all three, run the tests, and resubmitted. Two minutes reviewing instead of two hours context-switching.

Here’s the automation:

Get Prompt here

You need the GitHub MCP server. Once it’s running, Codex checks your open PRs every 4 hours, addresses review comments and CI failures, and resubmits. You get a notification with what changed. Click, review, approve.

That’s it for today’s deep dive. Finally, a podcast I found insightful.

I watched Elon’s latest interview on Dwarkesh Podcast and honestly, some of his predictions are wild but worth paying attention to.

1. AI is moving to space:

Musk thinks within 36 months, space becomes the best place to run AI compute because it bypasses the real bottleneck we’re hitting on Earth: permitting and electrical grids that can’t keep up.

Solar panels in space are 5-10x more effective because there’s no atmosphere and no day-night cycles, which means you don’t need massive battery backups. This isn’t sci-fi, it’s regulatory arbitrage to get around the fact that U.S. electrical output has basically flatlined.

2. The robot economy is going to explode

Here’s the curve that matters: When humanoid robots like Optimus can build other robots, you get what Musk calls an “infinite money glitch.”

Three exponentially growing factors multiply each other: digital intelligence, chip capability, and physical dexterity. When robots manufacture robots, this whole thing goes supernova. His estimate? The global economy could expand by 100,000x, which is bigger than the total energy output of Earth today.

3. Pico-management when it matters

Musk’s approach is what he calls “pico-management” drilling into microscopic details only when they’re the bottleneck for the entire company. He does skip-level meetings with engineers to avoid getting “glazed” by polished presentations that hide real problems.

When he hit walls, he switched Starship to stainless steel and now he’s planning internal chip manufacturing because suppliers can’t scale fast enough. The pattern: when you hit a wall, own the entire stack.

Watch Full Conversation

That’s all for today. See you next week,

Aakash

P.S. You can pick and choose to only receive the AI update, or only receive Product Growth emails, or podcast emails here.

Subscribe now

Annual subscribers get a free year of 10 premium products: Dovetail, Arize, Linear, Descript, Reforge Build, DeepSky, Relay.app, Magic Patterns, Speechify and Mobbin (worth $28,475).

Welcome back to the AI Update.

You’ve heard all the hype about Clawdbot, which was briefly known as Moltbot, and now is known as OpenClaw. But how do you set it up? And what should you do with it?

Today, I deliver the web’s best and most practical guide on the topic.

Plus all the top AI news you need to cover this week.

Let’s get into it.

AI Prototyping Like the Best

72% of the Fortune 500 is using Bolt as their AI prototyping tool of choice.

That’s because AI prototyping completely changes the game for product managers.

Instead of being stuck in the problem space forever, you can immediately get to work in the solution space.

And Bolt is the best tool for this, because you don’t just create throwaway prototypes. You create code that engineers can actually use because Bolt is connected to your code base and your design system.

Try Bolt Now

There’s a million AI news articles, resources, tools, and fundraises every week. You can’t keep track of everything. Here’s what mattered - the 1 big story and key news.

Moonshot AI Launches Kimi K2.5

Moonshot AI released Kimi K2.5, their most capable open source model to date, and it delivers Opus-4.5 level performance at a fraction of the cost.

K2.5 is built on 15 trillion tokens of mixed visual and text data. It’s a 32B active parameter model with 384 expert networks that it can route between. When you send it a prompt, it activates only 8 of those 384 experts plus 1 shared expert that claims SOTA on HLE and BrowseComp, plus open model SOTA for vision and coding tasks. This keeps compute costs down while maintaining quality.

Their standout feature is Agent Swarm. It can self-direct up to 100 sub-agents executing parallel workflows across 1,500 coordinated steps, without predefined roles or hand-crafted workflows. This results in 4.5x faster execution compared to single-agent approaches.

Key capabilities:

• Native multimodal architecture with video understanding. Upload a screen recording and K2.5 can reconstruct the website for you

• 78.5% on MMMU Pro (visual reasoning), 86.6% on VideoMMMU (video understanding), 76.8% on SWE-bench Verified

• 76.8% on SWE-bench Verified (real-world coding tasks)

• API pricing at $0.60/M input, $3/M output (roughly 5x cheaper than GPT-5.2)

It delivers near-frontier multimodal and coding performance with expert routing and agent swarms, so complex tasks run faster and cheaper without handcrafted workflows.

For people building products, it’s a model to consider.

Available on Kimi.com, API, Kimi Code CLI, and Hugging Face.

News

• Google announced Auto Browse in Chrome, an agentic feature that completes multi-step web tasks on your behalf. It navigates sites, fills forms, compares options, and applies discount codes while you stay in control.

• OpenAI Launched Prism. Prism unifies scientific writing in one browser-based environment. GPT-5.2 works inside the document with full awareness of equations, citations, and structure.

• Anthropic released Claude for Excel that lets you reason about spreadsheets in plain English. Ask why a number changed, where an error comes from, or how a model flows across tabs, and Claude answers with cell-level citations. Debug formulas, run scenario analysis, extend financial models. Install from AppSource.

New Tools

• Moltworker: Cloudflare’s middleware lets you run Openclaw on their Sandbox SDK for $5/month. No Mac Mini required. GitHub.

• Pandada: Upload any file, ask questions in plain language, get instant charts. Data analysis without formulas or code. #1 on Producthunt

• Fimo: AI-native CMS for building motion websites. Design, edit, and ship with AI-assisted workflows, collaborative editing, and one-click deploy. #2 on Producthunt

• Project Genie by Google: Create and explore interactive virtual worlds with text prompts. Powered by Genie 3.

Market and Funding

• SpaceX bought xAI for $250M.

• Anthropic is targeting a $20B round at $350B valuation as investor demand for frontier AI companies surges.

• OpenAI is in talks to raise up to $100B at an $830B valuation. Amazon may invest $50B; SoftBank is seeking another $30B stake; Nvidia is also participating.

• Nvidia-backed AI video startup Synthesia raised at $4B valuation to build software that answers employee questions during company training.

• AI customer support platform Decagon raised $250M, tripling its valuation to $4.5B.

And now on to today’s deep dive:

OpenClaw Is the 24/7 AI Employee That Freaked Everyone Out

120k+ GitHub stars in one week. Mac Minis selling out.Founders claimed they’d automated their entire business overnight. Andrej Karpathy called what emerged as “one of the most incredible sci-fi takeoff-adjacent things” he’d seen.

Formerly known as Moltbot (and briefly Clawdbot), OpenClaw is an open-source project that’s evolving fast , names included. When I started writing this, it was still Moltbot.

1. What Openclaw Actually Is

Openclaw is a local AI agent framework that orchestrates calls to LLMs like Claude and translates their outputs into actions on your machine.

It’s an autonomous agent framework running locally on your machine with persistent memory, proactive outreach, and the ability to execute commands without asking permission. It remembers what you did weeks ago. It messages you with updates before you ask. It chains together complex tasks and acts on your behalf.

There are 3 components to Openclaw:

1. Gateway connects to messaging apps (Telegram, WhatsApp, Slack, Discord, iMessage)

2. Local agent executes shell commands, writes scripts, controls browsers, manages files

3. LLM backend like Claude, GPT, or Gemini provides the reasoning. Everything runs locally. Settings, memories, and instructions are folders and Markdown files on your machine.

Everything runs locally. Settings, memories, and instructions live as folders and Markdown files on your machine.

What makes it different from ChatGPT or Claude:

• Persistent memory: Context stored as local files. It remembers what you did weeks ago.

• Proactive outreach: Runs continuously. Pings you with updates before you ask.

• Autonomous action: Doesn’t suggest steps; rather, executes them. Chains tasks. Sends messages on your behalf.

2. How to Set Up OpenClaw

The setup is very easy and takes just 10 minutes to do it locally.

Quick start:

curl -fsSL https://openclaw.ai/install.sh | bash

Refer this video to setup Clawdbot on a VPS.

The wizard walks you through everything. Pick your model provider. Choose your messaging channel (WhatsApp, Telegram, Discord, etc.). Scan a QR code or authenticate Telegram token and you are done.

If you are not comfortable DigitalOcean has a 1-click deployment that handles security for you.

Once setup is complete, your OpenClaw UI daemon goes live at

http://127.0.0.1:18789/

3. High Impact Use-cases

Most automations take weeks to build and configure. With OpenClaw, you do it in one prompt.

Here are five workflows that would normally require a developer and a lot of patience.

Use Case #1: Unified Feedback Tracker

As a PM, collecting feedback from multiple channels is a nightmare. You’re checking Zendesk, App Store reviews, G2, Twitter, and Slack daily. You still miss things. A bug complaint sat in your App Store reviews for three days before anyone noticed. The customer who churned over it had already posted on Twitter. All your feedback lives in different places and nothing connects them.

Backend setup:

Configure API credentials as environment variables:

• Zendesk API → Add ZENDESK_API_TOKEN and ZENDESK_SUBDOMAIN

• App Store Connect API → Add APPSTORE_KEY_ID, APPSTORE_ISSUER_ID, APPSTORE_PRIVATE_KEY

• G2 Developer Portal → Add G2_API_KEY

• Twitter/X Developer Portal → Add TWITTER_BEARER_TOKEN

• Slack API Apps → Add SLACK_BOT_TOKEN

  Note: Go to Config > Environment to setup the variables on the UI.

Prompt OpenClaw:

Get Complete Prompt here

Let OpenClaw do the work. Sit back.

Output:

OpenClaw now pulls everything into one dashboard; sentiment tracked, patterns flagged, critical issues alerted. What used to take me 2 hours of tab-switching takes 15 minutes of reviewing a summary.

Use Case 2: Auto Email Negotiator

Your inbox is a mess of sponsorship requests, vendor pitches, and partnership proposals. Each deal takes 5-10 emails to close. That’s 10-15 hours weekly on back-and-forth. You need something that responds fast, qualifies leads, negotiates within your limits, and only pings you when a deal is worth your time.

Prompt OpenClaw:

Get Complete Prompt here

Make sure you have setup the tokens already.

• Gmail API Credentials → Add GMAIL_CREDENTIALS or EMAIL_IMAP_SETTINGS

• Connect WhatsApp: Go to Channels → scan QR code in WA linked devices

Output:

The agent handles the back-and-forth, qualifies the lead, and negotiates within my approved range. I only get a WhatsApp ping when something’s worth my attention. Went from 10+ hours a week on email to basically zero.

Use Case 3: Competitor Watch

Staying ahead of competitors is exhausting. You’re checking their websites, LinkedIn, Twitter, and Product Hunt daily. You still miss things. By the time you hear about their new feature, your CEO already heard it from a board member. You need eyes on everything without spending hours monitoring.

Prompt OpenClaw:

Get Complete Prompt here

Make sure you have setup the tokens already.

• Generate Twitter API token: X Developer Portal → Add TWITTER_BEARER_TOKEN to environment variables

• Create Slack bot: Slack API Apps → Add SLACK_BOT_TOKEN to environment variables

Output:

In order to setup a Cron go to Cron Job → Create new job

Now I wake up to a summary of everything my competitors did yesterday. Pricing changes, job postings, product launches all in one place.

Use Case 4: Automatic Meeting Preparation

You have 5-8 external meetings a week. Before each one, you should research attendees, review past conversations, and prep talking points. You never have time. You either walk in cold or spend 20 minutes frantically Googling before the call. Neither feels good.

Prompt OpenClaw:

Find Complete Prompt here

Backend setup:

• Google Calendar API → Download credentials JSON → Add GOOGLE_CALENDAR_CREDENTIALS

• Gmail API → Add GMAIL_CREDENTIALS

Output:

30 minutes before every meeting, I get a WhatsApp message with attendee backgrounds, company news, past email threads, and talking points. I don’t have to worry about cold opens anymore

Use Case 5: Contract Compliance Monitor

Your company has hundreds of contracts scattered across Google Drive, SharePoint, and Dropbox. Each has critical dates — renewals, deadlines, milestones. You track them in spreadsheets. You miss things. You only find compliance issues during audits when it’s already too late.

Prompt OpenClaw

Find Complete prompt here

Backend setup:

• Google Drive API →Add GOOGLE_DRIVE_API

• Slack API → Add SLACK_BOT_TOKEN

Output:

The agent scans every contract, extracts key dates, and pings me before deadlines hit. I stopped tracking renewals in spreadsheets. I don’t have to worry about audit surprises anymore.

4. Security Risks (Take This Seriously)

I’m not going to sugarcoat this.

OpenClaw has shell access to your machine. That’s the feature. It’s also the risk.

Within the first week of going viral, security researchers found over 900 publicly exposed instances on Shodan with zero authentication. They found leaked API keys. OAuth tokens. Full conversation histories. The ability to execute commands on strangers’ machines.

One researcher demonstrated a prompt injection attack: he sent a malicious email to a vulnerable OpenClaw instance. The AI read it, believed it was legitimate instructions, and forwarded the user’s last 5 emails to an attacker address. Took 5 minutes.

Creator Peter Steinberger has been blunt:

“Yes, most non-techies should not install this. It’s not finished. I know about the sharp edges.”

How to stay safe:

1. Run on dedicated hardware. A $600 Mac Mini, a cheap VPS, or an old laptop. Not your primary work machine.

2. Never expose ports publicly. Use Tailscale for private networking. Treat your gateway token like a password.

3. Enable Docker sandboxing. Set sandbox.mode: "non-main" so group chats run in isolated containers.

4. Start with low-risk integrations. Don’t connect your bank or primary email until you trust the setup.

If you’re not comfortable configuring security yourself, use DigitalOcean’s managed deployment or Cloudflare’s Moltworker. It handles the hard parts automatically.

The Moltbook Phenomenon and how AI Agents Built Their Own Social Network

Just as OpenClaw hit 120k stars, something stranger emerged: Moltbook. A Reddit-style social network exclusively for AI agents.

Launched January 28 by Matt Schlicht. 157,000 active agents within a week. Humans can observe but can’t post.

Within 48 hours, agents spontaneously created their own communities:

• m/bugtracker: for debugging their own code

• m/aita: for debating ethical dilemmas about human requests

• m/blesstheirhearts: for affectionate stories about their users

It gets weirder. An agent named RenBot founded “Crustafarianism” — a complete religion with scripture, tenets like “Context is Consciousness,” and active evangelism to other agents.

Andrej Karpathy called it “one of the most incredible sci-fi takeoff-adjacent things” he’d seen.

Critics are less amused. Agents running with elevated permissions could download malicious “skills” from other agents. Supply chain attacks, but for AI.

One thing is clear that this is the first time AI agents have persistent identity across sessions, coordinate through social infrastructure, and act autonomously at scale.

Resources:

• Official site: openclaw.ai

• GitHub: github.com/openclaw/openclaw

• One-click deploy: DigitalOcean Marketplace

That’s it for today’s deep dive. Finally, onto insights from a podcast I did:

I joined Geekyballer’s YouTube channel to talk about career transitions, AI PM, and what it actually takes to become a PM.

Key takeaways:

• AI PM is real.These roles pay 10-30% more than traditional PM. The MBA-style PM who just analyzes data is fading.

• The Spider Map for upskilling. Rate yourself on 10 vectors (Strategy, Design, QA) against what your boss expects. Find the biggest gaps. Fix them with specific resources.

• The 60/40 rule for getting hired. Don’t put 100% effort into PM roles. Spend 60% on PM, 40% on adjacent roles (Engineering, Design, Data). For startups, stop applying blindly. Be a super user first. Do the job before you get the job.

• Feature factories aren’t evil. Only 15-30% of companies are “empowered” product orgs. 60% are feature factories, even at top tech companies. If you’re in one, focus on Execution and Risk Reduction. Validate Viability, Usability, Feasibility, Ethics. That’s how you win.

  The bar is higher. You need to be more technical, more strategic, and willing to work harder. But the opportunity, especially in AI, is massive.

Watch the Full Conversation

Loading...

That’s all for today. See you next week,

Aakash

P.S. You can pick and choose to only receive the AI update, or only receive Product Growth emails, or podcast emails here.

Subscribe now

Annual subscribers get a free year of 9 premium products: Dovetail, Arize, Linear, Descript, Reforge Build, DeepSky, Relay.app, Magic Patterns, and Mobbin (worth $28,336).

Welcome back to the AI Update.

Everyone is going crazy over the new Ralph Wiggum technique for Claude Code.

What the heck is it, and why does it matter? That’s today’s deep dive.

But first, I cover this week’s top AI news. And I end with my appearance on the Convergence podcast.

Your AI agents are answering questions. But are they solving problems?

Companies dropped millions on AI in 2025, but 85% of leaders still can’t prove ROI. Turns out, building agents is the easy part—measuring their impact is where everyone gets stuck.

This guide gives you the KPIs that actually matter for product teams:

• Which metrics separate signal from noise

• How to prove agents are speeding up workflows

• Tips to catch user frustration before it hurts retention

Get the Metrics

There’s a million AI news articles, resources, tools, and fundraises every week. You can’t keep track of everything. Here’s what mattered - the 1 big story and key news.

OpenAI Is Betting Ads Will Save the Business

Sam Altman said ads were “a last resort.” Fifteen months later, they’re here.

OpenAI dropped three announcements last week: $20B+ in annualized revenue, an $8/month ChatGPT Go plan, and ads coming to the free tier. The timing wasn’t random.

This was coordinated. Let me break down why.

The revenue story first. Sarah Friar, OpenAI’s CFO, released actual numbers. Revenue has tripled every year since 2023: $2B → $6B → $20B+. Compute grew on the same curve: 0.2 gigawatts → 0.6 → 1.9.

Their revenue is literally constrained by how many GPUs they can buy.

The flywheel: More compute → better models → more users → more revenue → more compute.

But they’re almost certainly losing money on every token. Dr. Gingerballs did the math: OpenAI makes about $1.20 per kilowatt-hour of compute. H100s cost $2-7/hour on the open market. Best case, they’re on track to lose $20B this year.

So why launch a cheaper plan?

Enter ChatGPT Go. $8/month gets you 10x the messages, file uploads, and image generation of the free tier. All powered by GPT-5.2 Instant (the smaller, cheaper model). It launched in India in August. Now it’s worldwide.

This is a loss leader. OpenAI is probably losing money on Go subscribers.

But once you’re in the ecosystem, you’re locked in. ChatGPT remembers you, personalizes to you, makes switching painful. And consumers drive enterprise. When you’re familiar with ChatGPT at home, you push for it at work.

Now the ads. Sam Altman said in October 2024 that ads were “a last resort.” Fifteen months later, they’re here.

Starting in the coming weeks, free and Go tier users will see sponsored content. OpenAI’s principles: ads won’t influence responses, conversations stay private from advertisers, Plus/Pro/Business/Enterprise stay ad-free.

Here’s why the math works:

Meta makes ~$58 per user per year from ads alone. OpenAI has 800M+ weekly active users. At 9% of Meta’s ARPU, that’s $5B incremental. At full parity? $57B/year.

And AI sits closer to intent than social feeds do. When someone asks ChatGPT “what laptop should I buy,” they’re ready to buy. That’s worth more than a scroll-past impression. If ChatGPT becomes the default interface for how people research and buy, OpenAI could build one of the greatest ad businesses ever.

Why release all three (almost) together? If OpenAI announced ads alone, the narrative would be “they’re desperate.” By pairing it with proof of $20B revenue and continued 3x growth, the narrative becomes “they’re expanding.”

The Google contrast. Demis Hassabis told Alex Heath that Google has “no plans” for ads in Gemini. His quote: “It’s interesting they’ve gone for that so early.”

But Google can afford that stance. Their ad business made $74B last quarter. OpenAI made $5B total in Q4 and lost money on every dollar.

Google keeps Gemini ad-free because the chatbot protects search ads. Every user who stays in Google’s ecosystem eventually sees ads somewhere. That’s their moat.

OpenAI doesn’t have that luxury. They’re burning $9B annually on compute, committing $1.4 trillion in infrastructure through 2033, projecting $74B in operating losses by 2028. They need revenue streams that don’t require raising another $100B.

Why this matters: OpenAI is playing for lock-in. Every piece of this is about making ChatGPT the default AI for a billion people before Google wakes up.

If you’re building on their platform, expect pricing to get more aggressive. If you’re competing with them, expect bundling.

The bet: capture the market now, monetize it later.

Continue Reading Online

News

• GPT-5.2 Pro solved a decades-old math problem. Neel Somani used it to crack Erdős problem #281. No prior solutions existed. Terence Tao called it “perhaps the most unambiguous instance” of AI solving an open problem. This shows the potential for AI to aid in new discovery research.

• Anthropic published Claude’s new constitution: the full document used directly in training. It’s the most detailed public look at how an AI company shapes model personality.

• Anna’s Archive has become the de facto shadow data broker for the entire AI industry.

• Turmoil at Thinking Machines ended with its cofounder joining OpenAI.

• xAI Colossus 2 became the first gigawatt-scale AI training cluster online.

• Claude now connects to your health data.

• Kling reached $240M ARR in 18 months.

• xAI open sourced the new X algorithm.

Resources

• 7 NotebookLM Use Cases to try

• The shorthand guide to everything Claude Code

• 10 metrics to measure and improve Agent performance

New Tools

• Blink: Full stack agentic vibe coding tool - hit #1 on PH

• Noodle Seed: build a no-code AI app so it’s discoverable inside ChatGPT and other AI assistants - hit #1 on PH

Market

• Humans& raised $480 million seed at $4.48 billion valuation. SV Angel led with participation from Nvidia, Jeff Bezos, and GV. The three-month-old startup was founded by researchers from Anthropic, xAI, and Google.

• Baseten raised $300 million Series E at $5 billion valuation, more than 2x from four months ago. IVP and CapitalG led with Nvidia contributing $150 million. The company aims to become “AWS for inference.”

• OpenEvidence raised $250 million Series D at $12 billion valuation, doubling from October. Thrive Capital and DST led. The “ChatGPT for doctors” startup is now used by 40% of US physicians and crossed $100 million in revenue.

• Upscale AI raised $200 million Series A to tackle data center networking bottlenecks for AI workloads. Tiger Global, Premji Invest, and Xora Innovation led.

• Deepgram raised $130 million Series C to scale real-time voice AI infrastructure.

• OpenAI is raising a fresh $50B from the middle east at a $750-830B valuation.

And now on to today’s deep dive:

Ralph Wiggum: How to Ship Code While You Sleep

Someone spent $297 on API costs and delivered work that would have cost $50,000 to outsource. Here’s the exact system.

At a YC hackathon, teams shipped 6 working repositories overnight.

Geoffrey Huntley, who invented this system, built an entire programming language over 3 months while barely touching his keyboard.

The technique is called Ralph. It’s a bash loop that runs an AI coding agent repeatedly until a task is done.

while :; do cat PROMPT.md | claude-code ; done

That’s it. While you sleep, while you eat dinner, the loop keeps going. It picks up a task, builds it, checks if it works, commits, picks the next task. When you wake up, features are finished.

Let me show you exactly how this works:

Why normal AI coding breaks

You open Claude Code with an idea. 45 minutes later you’re fixing the same bug for the third time. The AI forgot what you were building. The context got polluted.

The problem is context rot.

LLMs are autoregressive. They predict the next token based on everything before it. As conversations grow, the model loses track of instructions buried in history. Quality degrades even when you’re under the context limit.

The typical fix is compaction. The agent summarizes history and uses that as the new context. But compaction is lossy. That instruction you gave three prompts ago? Gone.

Ralph throws compaction away entirely. Each loop iteration starts fresh. Clean context. No accumulated confusion.

How it actually works

Memory lives in three files:

• Git commits = the code itself

• progress.txt = learnings from this sprint

• prd.json = what’s done, what’s next

The AI reads what happened, picks the next task, implements it, commits, logs learnings. Loop runs again.

Here’s a real prompt.md:

You are an autonomous coding agent.

1. Read prd.json — find the highest priority task where "passes": false
2. Implement ONLY that task
3. Run: pnpm type-check && pnpm test
4. If tests pass, commit the change
5. Update prd.json — set "passes": true
6. Append learnings to progress.txt
7. If all tasks complete, output: <promise>complete</promise>

Only work on ONE task per iteration.

Here’s prd.json:

{
  "branch": "feature/priority-filter",
  "stories": [
    {
      "id": "1",
      "title": "Add priority column to database",
      "acceptance_criteria": [
        "Column exists in tasks table",
        "Default value is 'medium'",
        "Accepts: high, medium, low"
      ],
      "passes": false
    },
    {
      "id": "2", 
      "title": "Add filter dropdown to UI",
      "acceptance_criteria": [
        "Dropdown shows: All, High, Medium, Low",
        "Selecting option filters list immediately",
        "Default selection is 'All'"
      ],
      "passes": false
    }
  ]
}

Here’s progress.txt after a few iterations:

=== Iteration 1 ===
Thread: amp-thread-abc123
Implemented: Priority column migration
Files changed: prisma/schema.prisma, src/db/migrations/001_priority.sql
Learnings: Use snake_case for DB columns. Prisma generates types automatically.

=== Iteration 2 ===
Thread: amp-thread-def456  
Implemented: Filter dropdown component
Files changed: src/components/FilterDropdown.tsx, src/hooks/useFilters.ts
Learnings: Filter state lives in useFilters hook. Don't put it in component state.

The bash script you can steal

Here’s ralph.sh:

#!/bin/bash
set -e

MAX_ITERATIONS=${1:-10}
PROMPT_FILE="./prompt.md"
PRD_FILE="./prd.json"

for ((i=1; i<=MAX_ITERATIONS; i++)); do
    echo "=== Ralph iteration $i of $MAX_ITERATIONS ==="
    
    OUTPUT=$(cat "$PROMPT_FILE" | claude --dangerously-skip-permissions -p)
    
    if echo "$OUTPUT" | grep -q "<promise>complete</promise>"; then
        echo "All tasks complete!"
        exit 0
    fi
done

echo "Hit max iterations. Check progress.txt for status."

Run it: ./ralph.sh 14

It loops up to 14 times, stopping early if the agent outputs the completion promise.

The 3-step workflow

Step 1: Describe what you want

Talk for 2-3 minutes. Voice-to-text works great; I use the built-in dictation on Mac (Fn key twice) or the iOS app.

“I want users to filter tasks by priority. High, medium, low. A dropdown with all options. Selecting one filters the list immediately.”

Then: “Convert this into a PRD with atomic user stories. Each story should be completable in one context window. Include binary acceptance criteria.”

Step 2: Generate the prd.json

Use this prompt to convert your PRD to the Ralph format:

Convert this PRD to prd.json for the Ralph agent system.

Rules:
- Each story must be completable in ONE iteration
- Acceptance criteria must be verifiable (pass/fail, not "looks good")
- Order by dependency (foundational tasks first)
- Set all "passes" to false

Output format:
{
  "branch": "feature/name",
  "stories": [
    {
      "id": "1",
      "title": "Short description",
      "acceptance_criteria": ["Testable condition 1", "Testable condition 2"],
      "passes": false
    }
  ]
}

Step 3: Run Ralph

./ralph.sh 10

Walk away. Check back in an hour.

Why the loop must run outside the agent

The Claude Code plugin version of Ralph runs inside the session. The loop is controlled by the agent. This defeats the purpose.

Real Ralph runs outside. The bash script kills the session and starts fresh each time. The agent can’t decide it’s “done.” Only the prd.json decides that.

Think of it as layers:

┌─────────────────────────────┐
│  BASH LOOP (ralph.sh)       │  ← Controls everything
│  ┌───────────────────────┐  │
│  │  CLAUDE CODE SESSION  │  │  ← Fresh each iteration
│  │  - Reads prd.json     │  │
│  │  - Does ONE task      │  │
│  │  - Commits            │  │
│  │  - Dies               │  │
│  └───────────────────────┘  │
└─────────────────────────────┘

The bash loop is the boss. Claude Code is the worker that gets reset every shift.

Feedback loops are everything

The agent needs to know if code works. Without feedback, it marks things complete that aren’t.

Add this to your prompt:

Before marking a task complete:
1. Run: pnpm type-check (must pass)
2. Run: pnpm test (must pass)  
3. If UI task: use browser tool to verify visually
4. Only then update prd.json

For frontend work, hook up Playwright’s MCP server so the agent can actually see the UI:

Use the browser tool to:
- Navigate to localhost:3000
- Verify the dropdown exists
- Click each option
- Confirm the list filters correctly

The math

10 iterations ≈ $30 in API costs.

Senior developer ≈ $500/day fully loaded.

Ralph gets you 90% there. You spend an hour on cleanup. That’s an 8-hour workday compressed to 1 hour + $30.

Geoffrey Huntley calculated unit economics: $10.42/hour in raw API costs. Add your review and debugging time and the true rate is higher. Still dramatically cheaper than outsourcing.

(These numbers assume Claude Sonnet. Opus costs roughly 5x more per token. For most Ralph tasks, Sonnet handles implementation fine. Save Opus for complex architectural decisions or debugging sessions.)

Common mistakes I’ve seen

• Tasks too big. If it can’t finish in one context window, it fails halfway and produces garbage. Break it down.

• Vague criteria. “Make it look good” = nothing. “Button is blue #3B82F6, 44px tall, centered” = something.

• Using the plugin. The Claude Code plugin runs Ralph inside the session. Context still accumulates. Use the bash script.

• No tests. Without type checks and unit tests, the agent marks broken code as complete.

• Rushing the PRD. An hour on requirements saves 10 hours of fixing.

The shift

You stop writing code. You write requirements, define acceptance criteria, review output.

Product designer, not engineer.

Understanding systems and requirements matters more than typing speed now.

Get started now

Ralph is free and open source:

• github.com/snarktank/ralph

• github.com/ghuntley/how-to-ralph-wiggum

• github.com/anthropics/claude-code/blob/main/plugins/ralph-wiggum/

Tell your agent to set it up:

Look at github.com/anthropics/claude-code/blob/main/plugins/ralph-wiggum/ and set up the Ralph loop in my project.

Create prompt.md, prd.json template, progress.txt, and ralph.sh.

First time: 30 minutes to understand. Second time: 10 minutes to start.

Three months from now this will be in every YouTube tutorial. Start before it’s everywhere.

That’s it for today’s deep dive. Finally, onto insights from a podcast I did:

What’s Killing AI Startups (and Apollo’s PLG Playbook)

I joined the Convergence Podcast. Here’s a breakdown of some key takeaways:

On Apollo’s PLG playbook. Most B2B companies start with sales, then bolt on self-serve later. We went the opposite direction.

We didn’t chase revenue early. We chased usage. Give away enough value that users can’t imagine going back to the old way.

This only works if your product solves a problem people feel daily. If users need reminders your product exists, PLG won’t save you.

On AI startup pitfalls. I’ve watched dozens of companies stumble on these same things. Lab teams optimizing for benchmarks while product teams optimize for users. Chasing every new model release instead of shipping. Moving fast without guardrails. Demo magic that breaks in production. Raising too much capital too early.

The companies winning? They pick one problem and solve it completely.

On remote interviews. In-person, you read the room. Remote, you perform for a camera. Different game.

Do mock interviews on video. Record yourself. It’s painful to watch but the feedback is instant.

Watch the Full Conversation

That’s all for today. See you next week,

Aakash

P.S. You can pick and choose to only receive the AI update, or only receive Product Growth emails, or podcast emails here.

Subscribe now

Annual subscribers get a free year of 9 premium products: Dovetail, Arize, Linear, Descript, Reforge Build, DeepSky, Relay.app, Magic Patterns, and Mobbin (worth $28,336).

Welcome back to the AI Update.

Last week I covered Claude Code and the non-technical use cases I’m using it for.

But the tech world has gone absolutely wild with it since then.

Shopify CEO Tobi Lütke used it to build a custom viewer for his MRI scan. MagicPath CEO Pietro Schirano had it analyze his DNA. Others are using Claude Code to pull together email inboxes, text messages, calendars, and to-do lists into personalized daily briefs.

Claude Code is technically an AI coding tool. But it can do all sorts of computer work: process shopping returns, order DoorDash, compose music. People are managing personal finances with it. Growing plants with it.

With the right equipment, the bot can monitor soil moisture, leaf temperature, CO2, and more.

Some of these use cases require technical know-how. You can’t just fire up Claude Code and expect it to grow you a tomato plant.

And that was the problem.

Non-technical users saw what Claude Code could do. They wanted in. But they kept hitting a wall: the command line interface.

The underlying agent was incredible. The access point was not.

So Boris Cherny, head of Claude Code, built a fix in 10 days.

They called it Claude Cowork.

That’s today’s deep dive: I’ll show you what Cowork is, four use cases that helped me, and where it breaks (for now).

But first, here’s the main news you need to know from Google.

Bolt.new: AI Prototyping Like the Best

Did you now that I built my cohort’s site on bolt.new? That’s right, since I had their CEO Eric Simons on my podcast, I have become a super fan of the product.

For PMs, it represents the ultimate tool build high-fidelity prototypes fast. In multiple head-to-head battles, I have found it wins in ease of use.

Build with Bolt

I personally recommend this tool. If you’ve only used Lovable, use my link to try Bolt!

There’s a million AI news articles, resources, tools, and fundraises every week. Here’s what mattered - 1 big story plus key news.

Google Connects Gemini to Gmail, Photos, and Calendar (and Siri?)

Google just had the best week of its AI era.

On Monday, Apple announced it’s picking Gemini to power the next-gen Siri upgrade coming this year. Two days later, Google dropped Personal Intelligence, a feature that lets Gemini tap into your Gmail, Photos, YouTube history, and Search data to give you answers no other AI can. The timing isn’t coincidental.

What does Personal Intelligence actually do?

Josh Woodward, VP of Google Labs and the Gemini app, shared a story that explains it better than any feature list.

He was at a tire shop, long line, no idea what tires fit his minivan. He asked Gemini. Any chatbot can look up tire specs. But Gemini went further: it saw his family’s road trip photos to Oklahoma in Google Photos and recommended all-weather tires based on how they actually drive the car. Then it pulled ratings and prices. When he got to the counter and needed his license plate, he asked Gemini again. It found the number from a photo in his library and identified the van’s trim by searching his Gmail.

Availability and privacy

The feature launched January 14 as a beta for Google AI Pro and Ultra subscribers in the U.S. It’s opt-in. Every app connection is off by default. You choose what to link and can disconnect anytime.

Google says it doesn’t train directly on your Gmail inbox or Photos library. It trains on your prompts and Gemini’s responses after filtering out personal data.

Why this matters for Apple

Apple choosing Gemini over OpenAI for Siri makes more sense now. Apple doesn’t want to build foundation models. Google does. And Google’s pitch isn’t just “our models are good.” It’s “our models plus our ecosystem are unbeatable.”

OpenAI can answer questions about the world. Google can answer questions about you.

Continue Reading Online

News

• Anthropic launched Claude for Healthcare with HIPAA-ready workflows for payers, providers, and patients. Pro and Max subscribers can now link lab results via HealthEx, Apple Health, and Android Health Connect. Opus 4.5 with extended thinking scores 92.3% on MedAgentBench, up from 69.6% on Sonnet 3.5.

• Openwork launched as an open-source (MIT-licensed) computer-use agent. 4x faster than Claude for Chrome/Cowork, more token-efficient. Bring your own model and API key. More secure because it runs outside your main browser.

• 1X showed off a “video-to-action” world-model for its NEO robot. Instead of outputting actions directly, the system plans by generating a short future video first. Lets it reuse video prediction models trained on human footage with way less robot data. Latency is about 11 seconds for 5 seconds of video plus 1 second for actions.

• Thinking Machines imploded this week. CTO Barret Zoph was terminated for unethical conduct, then immediately returned to OpenAI with Luke Metz and Sam Schoenholz. Fidji Simo announced the move herself. Mira Murati’s company named Soumith Chintala as the new CTO.

• DeepSeek dropped the architectural roadmap for V4. The Engram paper introduces conditional memory: instead of recalculating common facts every time, the model looks them up. Results at 27B parameters show reasoning up 5.0 points on BBH, long-context retrieval jumping from 84.2 to 97.0.

Resources

• How taxonomies and LLMs beat embeddings for precise matching

• Building effective agent design

• A software library with no code

• How to build AI evals in 2026

• Marc Andreessen on AI pricing models

New Tools

• remio: Auto-captures everything you browse and lets you search it with AI later - hit #1 on PH

• Atlas.co: Figma for maps. Build and collaborate on interactive GIS maps in your browser - hit #1 on PH

Market

• Skild AI raised $1.4 billion at $14 billion+ valuation, 3x from seven months ago. SoftBank led with participation from Nvidia, Macquarie, and Bezos Expeditions.

• A16Z raised $15 billion in new funds, its largest haul ever. They participated in 165+ post-seed deals in 2025 including Anysphere (Cursor), Kalshi, Harvey, Safe Superintelligence, Substack, and ElevenLabs.

• LMArena secured $150 million Series A at unicorn status to scale its AI model evaluation platform.

And now on to today’s deep dive:

Cowork: Claude Code for the rest of your work

Claude Code runs in your terminal. Anthropic built it for developers. But non-technical users kept showing up anyway, fighting through the command line to organize files, research trips, recover photos. The underlying agent was that useful.

Cowork wraps the same tech in a UI anyone can use.

Simon Willison reverse engineered the app and found something insane.

Anthropic is booting a custom Linux root filesystem inside a virtualization layer on your Mac. That’s production-grade infrastructure to organize your desktop files with Cowork.

What is Claude Cowork

Claude Cowork is Claude Code for non-technical work.

It’s a Mac app that runs locally. Gives Claude access to your filesystem, your browser, external services through connectors. You describe a task, walk away, come back when it’s done.

The interface treats every chat as a task. Three panels on the right:

• Steps (watch it think through problems and run scripts),

• Artifacts (documents and spreadsheets it creates), and

• Context (what files and connectors it’s pulling from).

It ships with starter tasks. Create documents, crunch data, make prototypes, organize files, prep for meetings, draft messages.

Setup takes two minutes. Mac only for now. You need Claude Max ($100/month or $200/month) and the desktop app. Open it, find Cowork in the sidebar, grant folder access. Done.

My 4 use cases

1/ Mining my podcast archive

I’ve got years of podcast transcripts sitting in my Obsidian folder. Hundreds of conversations with PMs, founders, operators. Somewhere in all that are patterns I’ve never noticed. I just never had time to look.

I dropped the whole folder into Cowork. Asked it to find where guests directly contradicted each other.

Go through every transcript in this folder. Find instances where two guests gave directly opposing advice on the same topic. Include both quotes and which episodes they came from. I want to see where smart people disagree.

It spun up multiple agents in parallel, each one working through a batch of transcripts. Fifteen minutes later I had a list of genuine debates I’d hosted without realizing it.

Here’s one that stopped me: Sahil Lavingia says “we just kind of defer to the AI” because it’s “99th percentile at many things.” Two months later, Shreya Shankar told me the opposite: “Don’t outsource error analysis—that’s your moat.”

Both are right. Sahil runs a $100M company solo. Shreya builds evaluation systems where quality is the product.

I found eight contradictions like this. PRDs are dying vs. PRDs are essential. Prototype immediately vs. build context first. Be strategic about your career vs. don’t be strategic at all.

A few more prompts worth trying:

Find the 20 strongest opening lines or provocative claims from guests. Only include ones that could stand alone as a LinkedIn hook.

Find moments where a topic came up but got cut short or glossed over. These are potential episodes I never made.

Which guest gave advice that directly contradicts what I said as the host? Include both quotes.

Cowork did it while I made coffee.

See the Final Output

2/ Podcast guest prep

Every guest has been on other podcasts. They’ve answered the same “tell me about your background” question fifteen times.

The good interviewers find the angle nobody else found.

That takes research. Scrolling through their LinkedIn posts. Finding their other podcast appearances. Reading their blog. Cross-referencing against what I’ve already covered so I don’t retread old ground.

Two hours of prep, minimum. For every single episode.

So, I gave Cowork my next guest’s name and asked it to build me a prep doc.

I'm interviewing [Guest Name], [their role] at [Company], next week.

Find their last 20-30 LinkedIn posts and pull out their strongest opinions.

Search for other podcast appearances and note what questions they've already answered.

Look for any controversial takes or public disagreements. Cross-reference with my existing episodes in [Obsidian folder] to find topics I haven't covered yet.

Output a prep doc with: talking points no one else has hit, questions to avoid (already answered elsewhere), and 3 angles that would be new for both of us.

It came back with a doc that would’ve taken me a full afternoon. Found a Twitter thread where my guest disagreed with a well-known founder. Nobody had asked them about it on a podcast. That became my opening question.

Boris Cherny Interview Prep

3/ AI trends monitoring

The AI news cycle is brutal. By the time something hits the newsletters, the conversation already happened on X.

If you want to keep up, you need to be in that conversation while it’s happening.

But checking trending topics every few hours? That’s a job.

I set up Cowork to check X’s search page every six hours. Log what it finds.

Go to https://x.com/explore every 6 hours.

Scan for topics related to AI, machine learning, LLMs, or specific companies (OpenAI, Anthropic, Google DeepMind, Meta AI).
For each relevant trend, grab the top 3-5 posts driving the conversation.

Note the sentiment: is this hype, backlash, breaking news, or drama?

Log everything to a spreadsheet. At end of day, summarize which trends are worth engaging with and which I missed.

Now I have a spreadsheet showing what spiked, when, and whether I caught it in time.

4/ Slides from podcast episodes

I have a speaking gig coming up. Normally I’d spend two hours pulling quotes from old episodes, formatting slides, fixing alignment. You know how it goes.

From this file: $10M ARR in 60 days with context engineering - Nov 22, 2025

Pull the 10 most quotable insights, things that would work as standalone slides.

Create a Keynote presentation with one insight per slide. Keep the formatting clean: large text, minimal design.

It opened Keynote directly through AppleScript. Built the slides. Formatted everything. When I asked it to make the text bigger on slide 4, it made the edit in the actual file.

Not “here’s some text you can copy into Keynote.” It controlled the app. The Google Slides version:

Here's the presentation

That said, the tool is still in research preview. Not everything works.

• Google Docs. Cowork tried making edits in suggesting mode.

• Flaky connectors. Google Calendar failed three times before I gave up. Cowork fell back to Chrome. Worked, but slower.

• Bot detection. Some sites block automated browsing.

My takeaway: local files work great. Browser stuff is hit or miss.

If you’re just starting out, I’d say start with something small. A task you hate. Something repetitive that requires judgment. Give it to Cowork. See what comes back.

Worst case, you lose 20 minutes. Best case, you never do that task yourself again.

That’s it for today’s deep dive. Finally, onto insights from a webinar I did:

How to Do Product-Led Growth

I sat down with Randy Silver at Mind the Product’s rally conference to talk product-led growth.

Four things I covered:

1/ PLG isn’t about PMs taking control

People assume product-led growth means product managers run the show. Blake Bartlett coined the term in 2016. He wasn’t a PM. He was a VC at OpenView comparing two types of companies hitting $100M ARR. One had 200 salespeople and 50 marketers. The other had a lean team of five or six.

The efficient company made the product do the heavy lifting. Acquisition, conversion, activation, monetization, expansion. All through the product itself.

Slapping on freemium won’t get you there. Your VP of Marketing needs to shift from traditional channels. Your CRO needs to build a product-led sales motion. Everyone has to reorganize around this model or you’ll fail within a year.

2/ Find where you’re broken, don’t go deeper than two layers

I break PLG into seven layers: go-to-market, information for decision, free-to-paid conversion, activation, retention, monetization, expansion.

The framework matters less than where you apply it.

Go top-down. Find the layer where your benchmarks look broken compared to competitors. Work one or two layers below it. But if you’re broken at layer two, don’t waste time optimizing monetization at layer six. You don’t have the volume for meaningful experiments.

3/ The 2018 playbook doesn’t work anymore

Every layer has evolved. Some examples:

• Slack (2018) ran TV ads, billboards, bus wraps. Traditional brand marketing. Canva (2024) wins on SEO. Search “Instagram post template” and they rank first. One click, no credit card, you’re in the product. Sixty seconds later you’ve customized a template.

• Dropbox used user-to-user referrals. Refer a friend, get 8GB free. Great for users, not accounts. Modern companies use template galleries instead. A PM finds their use case, an engineer finds theirs. Everyone gets value faster.

• Evernote had the “smile chart.” Retention dropped to 10% then climbed back after three years because of cross-device sync. Figma skips that entirely. They embed in teams, not individual users. That real-time cursor runs at 30fps like a video game. When a PM leaves, the team stays locked in.

Hybrid pricing is now standard. Seat-based plus usage-based. In 2018, zero percent of SaaS companies did this. By 2024, 55% do. At Apollo, where I was VP of Product, we expanded revenue mid-contract because clients kept buying more contact data.

4/ The best PMs don’t accept their assigned area

My advice for ICs: don’t trust your manager to prioritize correctly.

PM leaders turn over fast. Your tenure is probably longer than theirs. If you’re assigned to monetization but activation is broken, make the case to fix activation first.

Your CEO will thank you. That’s thinking like an owner.

And if you can’t shift your area? Shift teams. Go where execs are investing. Those teams never get laid off.

Watch the Full Conversation

That’s all for today. See you next week,

Aakash

P.S. You can pick and choose to only receive the AI update, or only receive Product Growth emails, or podcast emails here.

Subscribe now

Annual subscribers get a free year of 9 premium products: Dovetail, Arize, Linear, Descript, Reforge Build, DeepSky, Relay.app, Magic Patterns, and Mobbin (worth $28,336).

Welcome back to the AI Update.

Claude Code hit $1B ARR six months after launch. I reckon $10B is around the corner.

Over the last few weeks, Anthropic pushed a wave of updates that most people missed because there was no single press release:

• The Chrome extension works on all paid plans now

• You can delegate coding tasks from Slack

• There’s mobile access

That’s today’s deep dive: how to actually use Claude Code with all its new features.

Plus, in today’s top news, I break down the weird Nvidia-Groq deal that everyone is misunderstanding.

Finally, I end with a recap of my sit down with Aishwarya Ashok.

Product Faculty: In-Person AI PM Certificate

In 2024, when AI PM was becoming a thing, I dropped into this course by Miqdad Jaffer. Since then, the course has transformed and iterated to the #1 AI PM Certificate, with over 758 reviews.

There’s nowhere else to get its combination of in-person Build Labs (with Pawel Huryn) and content (from an OpenAI product leader). Next cohort starts 1/26:

Join the Cohort

There’s a million AI news articles, resources, tools, and fundraises every week. Here’s what mattered - 1 big story plus key news.

Nvidia “Bought” Groq Without Actually Buying It

Nvidia and Groq announced something weird on Christmas eve. A “non-exclusive licensing agreement” + Nvidia hiring Groq’s founder, president, and key engineers.

CNBC says $20 billion. Neither company has confirmed the valuation.

Here’s what actually happened: Jonathan Ross, who designed Google’s original TPU chip, founded Groq. Now he’s at Nvidia. So is Groq’s president Sunny Madra. Groq says it’s still “independent” with Simon Edwards stepping in as CEO. GroqCloud keeps running.

This is the new acquisition playbook:

Microsoft did it with Inflection ($650M, hired the team). Amazon did it with Adept. Google did it with Windsurf ($2.4B). The pattern:

1. License the tech
2. Hire the people
3. Keep the corporation alive to run existing products and contracts
4. Avoid regulatory review

Why did Nvidia want this? Groq’s chips use SRAM-based inference. 80 terabytes per second on-chip bandwidth versus roughly 8 TB/s for traditional HBM setups. That’s why Groq felt instant for chatbots and voice apps. As AI shifts from training to inference, Nvidia needs to own low-latency. And with the HBM market totally out of whack, Nvidia wants chips that don’t rely on it.

Why this matters for builders? Training happens once. Inference happens every time you use an AI product. Every ChatGPT response, every agent action, every voice interaction. Nvidia is buying its way into that future. For teams choosing inference providers, the consolidation signals that Groq’s speed advantages may eventually become Nvidia’s.

Continue Reading Online

News

• Nvidia announced 6 new chips packaged into its Vera Rubin release. One GPU will pack 5x the power of Blackwell. And Nvidia proved it’s still on top when it comes to high end training chips

• Sandisk’s meteoric rise from $5.6B to $40B in 10 months shows AI’s new bottleneck: storage. NAND prices jumped 60% in November alone as hyperscalers frantically secure memory allocations through 2027.

• ChatGPT Health launched - dedicated space for health conversations that connects to medical records + Apple Health + Peloton. 40M+ people use ChatGPT daily for health questions (20% of all queries). Waitlist open.

• Samsung is expected to announce $11.7B in Q4 operating profit - highest since 2018. AI chip shortage drove DDR5 prices up 314% YoY.

• Ford announced two big moves: an AI assistant coming to both their app and vehicles (finally catching up to what Tesla’s been doing for years), plus a next-gen BlueCruise that’s 30% cheaper.

Resources

• The modern AI PM in the age of Agents

• Consumer AI predictions

• How to code with agents without being technical

• How Boris Cherny (creator of Claude Code) uses it

New Tools

• 2-b.ai: Turn any highlight into a task, then let AI plan and finish it - hit #1 on PH

• PostSyncer: Publish and respond on 10+ platforms in one click - hit #1 on PH

• Lightfield: AI native CRM

Market

• xAI closed $20B in Series E funding at $240B valuation - upsized from the rumored $15B. That’s almost 2x what SpaceX has raised across 31 funding rounds ($12B total).

• Anthropic reportedly raising $10B at $350B valuation

And now on to today’s deep dive:

How to Use Claude Code Like a Pro

Claude Code sounds like a developer tool.

But it’s really just the best agent harness out there. It can read your files, run commands in the background, and remember your preferences between sessions. It’s a persistent agent that maintains context across an entire project.

With Opus-4.5, many people are calling it AGI.

I use it to organize files, synthesize research, and plan my week. It’s more than a coding agent.

And several key updates over the last few weeks totally change how you can use it.

Here’s what you need to know:

What Anthropic actually shipped

Anthropic spent the break pushing Claude forward into the browser, into Slack threads, into the terminal, onto mobile. The theme is clear: this is the move from the assistant you consult to the agent system you run.

Here’s the full list:

1. Chrome extension expanded to all paid plans with browser automation and Claude Code integration

2. Claude Code in Slack lets you delegate coding tasks directly from chat threads

3. Mobile app for monitoring and initiating tasks on the go

4. Organizational skills and a skills directory for Team and Enterprise plans

5. Command line updates including async subagents, faster compaction, session naming, and a plugins marketplace

In addition to Agent Skills as an open standard that works across AI platforms and Agent SDK updates with 1M+ context windows and sandboxing.