Skip to content
Aakash GuptaAakash Gupta

He Uses 7 Claude Code Agents to Build Apps with 0 Employees

Gabor Mayer is a PM at Google who runs a 21-agent Claude Code development team. In this episode, he walks through a live demo building a production mobile app from zero to TestFlight - Confluence for specs, JIRA for tickets, Figma for design, and Claude Code for development. Full Writeup: https://www.news.aakashg.com/p/claude-code-dev-team --- Timestamps: 00:00 AI agents can now run a startup workflow 01:23 Subscribe and AI tools bundle 01:55 Claude Code as your designer, developer, and systems analyst 02:43 Gabor's 21-agent startup team inside Claude Code 04:57 Inside the system analyst agent 05:52 Live demo: zero to TestFlight 08:42 Prompting Claude to define a good system analyst 10:02 Ads 11:53 Building the system analyst workflow 12:17 Why documentation matters: Confluence, Jira, and MCPs 15:30 Why classic PM skills make you a better AI builder 19:15 The scaffolding that prevents AI spaghetti code 22:17 Setting up project-specific agents in Claude Code 26:19 Dictating the full product spec for the hockey rules app 32:19 Ads 35:29 Why dictation changes the quality of AI specs 47:30 Creating the visual direction in Figma Make 55:59 The idea-to-prompt-to-design-to-app workflow 1:06:21 Claude Code starts building the Figma screens 1:23:59 Frontend epics and Figma-linked tickets appear in Jira 1:48:49 The hockey rules AI app is live 1:53:56 Full recap: Claude, Confluence, Figma, Jira, Simulator, TestFlight 2:03:17 Should PMs get AI PM certificates? 2:08:15 How to create a PM portfolio that helps you land top jobs 2:13:32 How to get started building with AI agents --- 🏆 Thanks to our sponsors: 1. Maven: Go from PM to AI builder with Claude Code - https://bit.ly/4bPulv7 2. Amplitude: The market-leader in product analytics - https://amplitude.com/session-replay?utm_campaign=session-replay-launch-2025&utm_source=linkedin&utm_medium=organic-social&utm_content=productgrowthpodcast 3. Testkube: Leading test orchestration platform - http://testkube.io/ 4. Land PM Job: 12-week experience to master getting a PM job - https://www.landpmjob.com/ 5. Product Faculty: Get $550 off their #1 AI PM Certification with code AAKASH550C7 - https://maven.com/product-faculty/ai-product-management-certification?promoCode=AAKASH550C7 --- Key Takeaways: 1. One-prompt vibe coding fails because of context compression - When you give one agent one massive specification, the model silently drops details it considers lower priority. Your color palette, edge cases, and security requirements disappear. Break work into smaller scoped tasks with dedicated agents. 2. The system analyst agent is the most important agent in any AI dev team - It asks clarifying questions one at a time, documents decisions in Confluence, and maps dependencies before code is written. Without it, every agent operates on partial context. 3. Dictation produces 5x more specification detail than typing - Use voice tools like Super Whisper to describe your app requirements. Even imperfect dictation captures more nuance than careful typing. The AI handles the interpretation. 4. Reusable agents encode institutional knowledge - Every painful lesson, API workaround, and MCP quirk gets saved in the agent markdown file. The next project starts from a position of strength rather than from zero. 5. Attach screenshots to every front-end development ticket - Without visual references, coding agents default to generic AI aesthetics. The Figma link or screenshot is what ensures your brand design actually shows up in the code. 6. Build a Spaghetti Agent for code quality - A dedicated code maintainability agent checks naming conventions, circular references, and comment quality after every sprint. It catches structural problems a PM would never spot. 7. The coding phase is the fastest part of building - Specification, documentation, design, ticket creation, and team review take longer than the actual code generation. Do not skip the front-end work. 8. Sprint organization with dependency mapping is essential - Use tags as a workaround for Atlassian MCP limitations. Map dependencies between tickets so agents build features in the right order. Without sprints, agents build on top of code that does not exist yet. --- 👨‍💻 Where to find Gabor Mayer: LinkedIn: https://www.linkedin.com/in/mayergabor/ Maven Course: https://maven.com/gabor/productbuilder X: https://x.com/gabor_pm 👨‍💻 Where to find Aakash: Twitter: https://www.x.com/aakashg0 LinkedIn: https://www.linkedin.com/in/aakashgupta/ Newsletter: https://www.news.aakashg.com #claudecode #aipm --- 🧠 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.

Aakash GuptahostGabor Mayerguest
Apr 30, 20262h 15mWatch on YouTube ↗

CHAPTERS

  1. 0:00 – 2:43

    AI agents running a full startup workflow in hours

    Aakash frames the episode around a new reality: agents can now handle core startup tasks—PRDs, design, ticketing, coding, and shipping—end to end. Gabor introduces the promise: going from zero to an iOS TestFlight build in a single session using an agent “team.”

  2. 2:43 – 4:57

    The 21-agent “company” inside Claude Code (roles and responsibilities)

    Gabor explains how he models Claude Code agents as a real cross-functional org, expanding from 15 to 21 specialized roles. The system analyst is the central node, but he also uses agents for CTO/architecture, brand, design, testing, performance, privacy/data governance, and maintainability.

  3. 4:57 – 8:42

    Under the hood: the System Analyst agent definition and why it’s the linchpin

    They open an agent markdown definition to show how the system analyst breaks down requirements, flags ambiguity, asks clarifying questions, and tracks dependencies. Gabor emphasizes that high-quality documentation and ticketing originate here, improving downstream build quality.

  4. 8:42 – 11:53

    Kickoff from the consumer Claude app: voice-first ideation and role prompting

    Instead of starting in a terminal, Gabor begins in the Claude consumer app to make the workflow accessible and voice-friendly. He demonstrates prompting Claude to behave like a “good system analyst” and sets the app idea: an AI hockey rules assistant.

  5. 11:53 – 12:17

    Defining a “good system analyst” and the clarifying-questions-first workflow

    Gabor shows how he first asks Claude to define the system analyst role (good vs bad), then uses that to set expectations. He instructs the agent to ask questions one at a time and not to write docs until understanding is complete—preventing premature output and messy direction changes.

  6. 12:17 – 22:17

    Documentation + ticketing backbone: Confluence/Jira via MCP and why PM skills still matter

    Gabor connects Atlassian tools via MCP so Claude can write Confluence docs and create Jira issues directly. He argues classic PM skills (requirements, decomposition, documentation) are even more valuable with agents because they prevent unmaintainable “AI spaghetti code.”

  7. 22:17 – 26:19

    Project setup in Claude Code: agents, permissions, and guardrails

    Gabor switches to Claude Code, differentiating “acting like an agent” in chat vs real project agents in Claude Code. A live hiccup forces him to recreate the project folder and re-run setup, highlighting practical guardrails: watch permissions, keep work inside the project directory, and avoid risky access requests.

  8. 26:19 – 47:30

    Dictating the full product spec: stack, RAG sources, limits, privacy, and secrets

    Gabor dictates a detailed spec for “Rule Ask”: Flutter iOS app, Firebase backend, IIHF rulebook + situation book as RAG sources with vector embeddings, and a friendly referee persona. He adds cost controls (20k-word limit with 24h cooldown), privacy constraints (no server-side user storage), and strict API key handling via Firebase Secret Manager.

  9. 47:30 – 1:06:21

    Visual direction in Figma Make: brand kit, typography, and palette from inspiration

    Gabor uses Figma Make to generate a comprehensive design system (typography, colors, buttons, states) using inspiration screenshots. He notes an important prompt constraint—ask for “inspiration, not copying”—to avoid tool refusal and keep the workflow moving.

  10. 1:06:21 – 1:23:59

    Claude Code drives Figma: generating screens and a clickable prototype with a UX agent

    With MCPs connected (Figma, Chrome DevTools), Claude Code creates the actual Figma screens from the spec and style guide, then adds prototype links (“arrows”) automatically. They observe that agent-driven Figma manipulation produces usable UI quickly and avoids generic ‘AI-looking’ defaults when design context is correctly anchored.

  11. 1:23:59 – 1:48:49

    From design to execution: Jira epics/tickets with Figma links, sprints-by-tags, and parallel work

    The system analyst and other agents generate Jira epics and detailed tickets, ensuring every frontend ticket includes a screenshot or Figma link so developers don’t drift into generic UI. Backend setup tickets are created first (Firebase basics, domain, secrets), then frontend and backend workstreams parallelize; sprints are approximated using tags due to MCP limitations.

  12. 1:48:49 – 2:13:32

    Shipping the hockey rules app: simulator demo, RAG transparency, and TestFlight upload

    After sprint execution, the app runs in the iOS Simulator with a working RAG pipeline and an “observer mode” showing retrieval hits, token usage, and latency. They upload the first build to TestFlight, discuss Apple processing/review timelines, and recap the full pipeline from dictation to deployment.

  13. 2:13:32

    Career layer: AI PM certificates vs building, portfolios that prove capability, and getting started

    The conversation shifts to careers: Gabor argues certificates matter less than hands-on skill and proof of building. He recommends creating portfolio apps with demonstrable “stories” (debugging retrieval, tuning scoring/thresholds) and advises newcomers to start by asking questions in their preferred LLM, then accelerate with structured programs if needed.

  14. Auto-generating Confluence documentation from the clarified spec

    After questions are resolved, Claude writes structured Confluence pages (product overview, architecture, agent specs) via MCP. Gabor checks the docs primarily for fidelity—whether the generated artifacts accurately reflect the spoken requirements and technical approach.

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