Skip to content
Aakash GuptaAakash Gupta

How this PM Used Claude Code to Support 20 People

Hannah Stulberg is a PM at DoorDash and former Google APM with 1,500+ hours in Claude Code. She walks through her Team OS live - a shared repo where every function self-serves context without asking the PM. Full Writeup: https://www.news.aakashg.com/p/hannah-stulberg-podcast Transcript: https://www.aakashg.com/hannah-stulberg-podcast/ Team OS Repo: https://github.com/in-the-weeds-hannah-stulberg/team-os-example-repo --- Timestamps: 0:00 - Intro 1:45 - What is a Team OS 3:50 - Live folder walkthrough 6:34 - Context management theory 8:27 - Nested CLAUDE.md files 11:36 - Ads 13:37 - Shared skills and commands 17:24 - Scaling analytics 25:24 - Shared automations 31:10 - Ads 33:32 - Plan mode for docs 49:47 - Parallel agents 59:50 - The learning flywheel 1:04:22 - Which AI tool when 1:09:11 - Outro --- 🏆 Thanks to our sponsors: 1. Bolt - Ship AI products 10x faster - https://bolt.new/solutions/product-manager?utm_source=Promoted&utm_medium=email&utm_campaign=aakash-product-growth 2. Jira Product Discovery - https://www.atlassian.com/software/jira/product-discovery 3. Kameleoon - Prompt-based experimentation - http://www.kameleoon.com/ 4. Amplitude - Product analytics leader - https://amplitude.com/session-replay?utm_campaign=session-replay-launch-2025&utm_source=linkedin&utm_medium=organic-social&utm_content=productgrowthpodcast 5. Product Faculty - $550 off AI PM Cert with code AAKASH550C7 - https://maven.com/product-faculty/ai-product-management-certification?promoCode=AAKASH550C7 --- Key Takeaways: 1. Build a Team OS, not a personal OS - A shared repo where every function checks in work. Engineers, designers, and analysts self-serve without asking the PM. 2. Root CLAUDE.md is everything - Doc index, team roster with Slack IDs, channel map. Keep under one page or you burn context every session. 3. Nested indexes save 97% of context - Every folder gets a navigation CLAUDE.md. A customer query used only 3% of the context window. 4. Three token tiers - Always-loaded root (~500 tokens), folder indexes on navigation (200-500), content files on demand (1,000-10,000+). 5. Split analytics by product area - Metrics, queries, schemas separated. Progressive loading prevents waste. 6. Gate launches on repo updates - Feature not shipped until metrics, queries, schemas, and playbooks are checked in. 7. Verified playbooks kill hallucinations - Analyst-audited methodology. Claude follows verified steps instead of inventing its own. 8. Plan mode makes 10x docs - Shift+Tab twice. Five phases: load context, ask questions, build plan, push thinking, review agents. 9. Split long docs across parallel agents - Each writes to a temp file. Orchestrating agent compiles. Prevents context overflow. 10. The flywheel compounds daily - Automate one task, free time, improve the repo. After 1,500 hours still iterating every day. --- 👨‍💻 Where to find Hannah Stulberg: LinkedIn: https://www.linkedin.com/in/hannah-stulberg/ Substack: https://hannahstulberg.substack.com 👨‍💻 Where to find Aakash: Twitter: https://www.x.com/aakashg0 LinkedIn: https://www.linkedin.com/in/aakashgupta/ PM Newsletter: https://www.news.aakashg.com AI Newsletter: https://www.aibyaakash.com/ #claudecode #teamos --- 🧠 About Product Growth: The world's largest podcast focused solely on product + growth, with over 200K+ listeners. 🔔 Subscribe and turn on notifications to get more videos like this.

Hannah StulbergguestAakash Guptahost
Apr 6, 20261h 10mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Building a Team OS with Claude Code for PM leverage

  1. The episode introduces a “Team OS,” a Git-backed knowledge base designed so coding agents can retrieve the right context on demand without bloating the LLM session.
  2. Hannah explains context-management fundamentals—context window, compaction, and “thinking room”—and shows how nested, lean CLAUDE.md doc indexes prevent unnecessary context loading.
  3. The team standardizes reusable “skills,” commands, and workflows so unstructured inputs (like customer calls) become consistent artifacts that Claude can synthesize reliably.
  4. Analytics is scaled by storing metric definitions, vetted SQL queries, and table schemas in structured folders, reducing hallucinations and enabling PMs/engineers to self-serve analysis.
  5. High-quality docs come from deliberate planning using Plan mode, checkpoints, verification criteria, parallel agents writing to temp files, and saving plan files to avoid rework and “context rot.”

IDEAS WORTH REMEMBERING

5 ideas

Treat team context like a version-controlled product, not scattered docs.

By keeping PRDs, research, analytics references, and workflows in a repo, teams create shared, searchable context that an AI agent can use consistently across roles.

Keep the root CLAUDE.md extremely lean and push detail downward.

The root file loads every session, so it should include only high-frequency essentials (doc index, team roster/handles, key channels) while nested CLAUDE.md files act as local indexes.

Doc indexes reduce token burn and improve reasoning quality.

Nested indexes let Claude navigate directly to relevant folders, conserving context window and preserving “thinking room,” which improves reasoning compared with broad repo exploration.

Separate “summary” artifacts from “raw” artifacts to maximize fidelity.

Storing call summaries in a consistent format allows fast synthesis across many meetings, while transcripts remain available only when deeper detail is needed.

Standardized skills turn messy human input into machine-friendly structure.

Team-wide templates for things like customer call summaries create uniform outputs, enabling reliable cross-customer analysis even when many people contribute.

WORDS WORTH SAVING

5 quotes

I have spent now, like, 1,500 hours in Claude, and I'm still iterating on my setup and improving it literally every single day.

Hannah Stulberg

You don't want very much in your CLAUDE.md file. CLAUDE.mds should be very, very, very lean.

Hannah Stulberg

The whole repository is structured around helping Claude read and use the right information at the right time.

Hannah Stulberg

When we're rolling out a new feature, the feature is not rolled out until the repository is updated.

Hannah Stulberg

Claude is like a really, really eager and highly talented junior employee.

Hannah Stulberg

Team OS (team knowledge repo) conceptRoot and nested CLAUDE.md doc indexesContext window, compaction, and thinking roomStructured customer research: summaries vs transcriptsShared skills/commands and standard templatesAnalytics scaling: metrics, SQL queries, schemas, playbooksPlan mode, verification, parallel agents, saved plan filesGitHub PR workflow for non-technical teammatesMCP/CLI integrations (Slack, Google Docs, Snowflake, Playwright)Beginner mindset and continuous iteration

High quality AI-generated summary created from speaker-labeled transcript.

Get more out of YouTube videos.

High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.

Add to Chrome