Behind the Craft by Peter Yang

Behind the Craft by Peter Yang

Podcast

Inside Claude Code: How an AI Native Team Actually Works | Cat Wu

Claude Code product lead on getting user feedback every 10 minutes, choosing prototypes over docs, and why designers and PMs should check in code directly

Peter Yang's avatar
Peter Yang
Sep 14, 2025
∙ Paid
17
2
Share

Dear subscribers,

Today, I want to share a new episode with Cat Wu.

Cat is the product lead for Claude Code and she showed me how one of the most AI-native teams in tech actually works. Highlights include getting user feedback every 10 minutes, going straight to prototypes instead of docs, and designers checking in code directly. Cat also shared her 3 best Claude Code tips and what’s next for the product.

Watch now on YouTube, Apple, and Spotify.

Cat and I talked about:

  • (00:00) The process behind the best Claude Code features

  • (05:00) How engineers can own features from idea to launch

  • (10:04) Getting feedback every 10 minutes from 1000+ users

  • (14:00) How designers and PMs can ship production code

  • (19:58) How the Claude Code team does AI evaluations

  • (29:04) 3 tips from Cat to get the most out of Claude Code

  • (30:12) Why your Claude.md file is so important

  • (33:32) What's next for Claude Code and AI agents


Thanks to Miro for sponsoring today’s newsletter

Miro got to the bottom of what’s causing workflow bottlenecks in their new survey of 6,000+ knowledge workers worldwide, and the results aren’t pretty:

  • 🔴 62% say busywork drains their energy and 58% report reduced creativity

  • 🔴 83% push their best work outside the workday

One stat that offers encouragement: 76% believe AI can help reduce silos and cut repetitive tasks so people can focus more on creative, strategic work.

Read the full survey here


Top 10 takeaways I learned from this episode

  1. Ship demos, not docs. “We don’t use Google Docs much on our team. The source of truth is the code base.” Cat and the team prefer using Claude Code to prototype features to get user feedback ASAP instead of writing multi-page specs.

  2. Engineers own features end-to-end. "Many of our best features came from an engineer prototyping an idea and shipping it to internal users.” You can tell whether a feature is good or not by whether internal users and beta customers start using it organically. Just ship progressively and see what users love.

This post is for paid subscribers

Already a paid subscriber? Sign in
© 2025 Peter Yang
Privacy ∙ Terms ∙ Collection notice
Start writingGet the app
Substack is the home for great culture