Skip to content
Aakash GuptaAakash Gupta

The Claude Setup That Let a PM Beat 30 Engineering Teams

Jyothi Nookula used one idea from an Anthropic blog post to beat 30 engineering teams at an internal hackathon. In this episode she rebuilds that same system live, then walks through the entire Claude stack a modern PM runs, ending with an AI chief of staff that quietly learns your whole org. Full Writeup: https://www.news.aakashg.com/p/the-complete-claude-stack-for-pms Transcript: https://www.aakashg.com/jyothi-nookula-claude-stack-podcast/ --- Timestamps: 0:00 - Intro 2:47 - How Jyothi won a hackathon against 30 engineering teams 4:30 - The five-layer Claude stack overview 7:17 - Which model to use, Haiku vs Sonnet vs Opus 9:55 - Which surface to use, Chat vs Desktop vs Mobile vs Chrome 10:50 - Ads 14:53 - Cowork automations, morning brief, standup, end of day 26:08 - Building skills that beat prompts 34:28 - Building an AI chief of staff from scratch 59:24 - MCPs and integrations every PM needs 1:02:16 - Claude Design for decks and prototypes 1:08:39 - The hackathon reveal, building the adversarial agent 1:13:01 - The new AI builder role and changing PM ratios 1:28:51 - The self-improving product loop 1:32:04 - Outro --- Thanks to our sponsor: Hyper Agent: Turn your recurring PM work into reusable agents - https://hyperagent.com/productgrowth --- Key Takeaways: 1. Match the model to the job - Sonnet handles ninety percent of PM work at the best cost. Save Opus for genuinely hard reasoning, and hand fast bulk jobs to Haiku. Defaulting to the smartest model for everything just burns time and money. 2. Stop skipping the knowledge layer - Projects, skills, and memory are what make Claude know your actual work instead of guessing from a blank slate. Almost everyone underinvests here, and it is the difference between a chatbot and an assistant that knows you. 3. A skill beats a prompt - A skill is a saved playbook Claude picks up on its own when it fits the task. It only loads when needed, so it never clogs the context window. Build one once and stop re-explaining the same task forever. 4. Write your skills yourself - Human-written skill files consistently beat AI-written ones. Draft with Claude to move fast, then layer in the domain knowledge only you have. That last step is what makes it actually work. 5. Automate your time-based work - A morning brief, a standup summary, and an end-of-day wrap can all run on a schedule while you sleep. You walk in already knowing what needs your attention. It clears the busywork that eats your mornings. 6. Give your automations guardrails - Cap the length, tell them to stick to facts, and never let them hallucinate. Left unchecked, an AI brief will pad itself and invent things. A few hard rules keep it sharp and trustworthy. 7. Build a chief of staff that learns your org - Point Claude at your meeting notes and let it build a picture of your people, your priorities, and your politics over time. Feed it transcripts first, since they carry the richest signal. It compounds into something no generic chatbot can match. 8. Keep that knowledge base on your own laptop - Your most personal work data does not belong in someone else's cloud. When you leave a company, it walks out with you. You keep full control of your most sensitive context. 9. The PM job is changing fast - The ratio is shifting from one PM per eight engineers toward two PMs per one. Building is becoming part of the role, and the PMs who can ship are pulling ahead. 10. Building is easy now, taste is scarce - When anyone can build, the edge moves to knowing what is worth building and what good actually looks like. That judgment is the one skill you cannot download. --- Where to find Jyothi Nookula: LinkedIn: https://www.linkedin.com/in/jyothinookula/ NextGen Product Manager: https://nextgenproductmanager.com/ 👨‍💻 Where to find Aakash: Twitter/X: https://x.com/aakashgupta LinkedIn: https://www.linkedin.com/in/aagupta/ Newsletter: https://www.news.aakashg.com #claude #productmanagement #aipm --- About Product Growth: The world's largest podcast focused solely on product + growth, with over 200K+ listeners. Subscribe and turn on notifications.

Jyothi NookulaguestAakash Guptahost
Jun 30, 20261h 33mWatch on YouTube ↗

At a glance

WHAT IT’S REALLY ABOUT

Five-layer Claude stack helps PMs automate, build, and win hackathons

  1. Jyothi explains a five-layer “Claude stack” (models, surfaces, knowledge base, integration fabric/MCP, agents) as a practical framework for choosing the right Claude capabilities for each PM job.
  2. She details when to use Haiku vs Sonnet vs Opus, emphasizing Sonnet as the default for most PM work and noting Opus’s occasional “stuck” behavior despite stronger reasoning.
  3. She demonstrates Claude Desktop’s Cowork automations to produce morning briefs, standup briefs, and end-of-day summaries by connecting Calendar, Gmail, Drive, and Jira with strict formatting and anti-hallucination rules.
  4. She shows how to build a local, privacy-preserving “AI chief of staff” knowledge base using Claude Code in VS Code, then expose it via an MCP server so Claude Desktop can read/write organizational context over time.
  5. She recreates the hackathon-winning concept: a GAN-inspired adversarial agent loop that red-teams an agent, scores failures, iteratively hardens the system prompt, and stops when it meets a defined threshold.

IDEAS WORTH REMEMBERING

5 ideas

Treat Claude as a stack, not a single chatbot.

Jyothi’s five-layer model helps PMs diagnose what’s missing (wrong model, wrong surface, no context/KB, missing integrations, or weak agent orchestration) instead of endlessly tweaking prompts.

Default to Sonnet for most PM tasks; escalate only when needed.

Sonnet is positioned as the best quality-to-cost workhorse for PRDs, synthesis, and briefs; Haiku is for high-volume triage; Opus is for high-stakes reasoning but may loop or get “stuck,” so switching back can be pragmatic.

Picking the right surface is a leverage skill for PMs.

Web chat is fast for Q&A, Desktop is best when local files + Cowork automations matter, Mobile supports “on the go” monitoring, and the Chrome plugin enables computer-use workflows like competitive research or agent-driven user testing.

Cowork automations can replace brittle no-code workflows with natural language instructions.

Instead of box-and-arrow tools that fail when one step breaks, Cowork lets PMs schedule briefs and dashboards that pull from connectors (Calendar/Gmail/Drive/Jira/Slack), with guardrails like word limits and “never invent action items.”

Skills are reusable playbooks that reduce context bloat and improve consistency.

Multi-file skills load progressively (description first, detailed steps only when invoked), enabling richer SOPs like customer-interview synthesis with citations, evidence rules, templates, and links to frameworks such as Jobs-to-be-Done.

WORDS WORTH SAVING

5 quotes

Understanding which surface to reach for which use case, um, becomes one of the core PM skills that will help you become 10X more effective.

— Jyothi Nookula

Now, this is the layer that I think most PMs under-invest in. It's this layer that makes Claude go from being a generic chatbot to actually knowing your context.

— Jyothi Nookula

Writing code has become so easy now, right? So building is easy. It is thinking about the new capabilities and how you want to go solve the problems that your customers have.

— Jyothi Nookula

It's like making and building is easy now. Taste is what is important for us to develop.

— Jyothi Nookula

Previously, if you see one product manager, um, works with eight engineers, now it's like two product managers, one engineer.

— Jyothi Nookula

Five-layer Claude stack frameworkModel selection: Haiku vs Sonnet vs OpusSurfaces: web chat vs desktop vs mobile vs Chrome computer-useCowork scheduled automations (briefs, dashboards)Skills as progressive-disclosure playbooks (multi-file)Knowledge base + local MCP server (“chief of staff”)Adversarial agents and self-improving evaluation loopsClaude Design for decks, carousels, prototypesBuilder PM / AI builder role and changing PM:Eng ratios

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.

iOSAndroidClaudeChrome