By Teresa Torres · producttalk.org · @ttorres on X · LinkedIn
Teresa Torres and Petra Wille discuss how they use AI as strategic thought partners across writing, coding, and product work rather than as ghostwriters. Petra describes a voice-first, bilingual transcription workflow that lets her brainstorm and mine her own archive for consistency. Teresa explains how Claude Code in the terminal, with planning mode and local context, changed her coding and blog drafting, including reviewer sub-agents that critique her drafts. They compare ChatGPT and Claude strengths, stress that they still write every word themselves, and reflect on English becoming the interface for personal agents. The conversation matters because it offers a practical model for keeping human authorship while using AI for critique and thinking.
01Key takeaways
- Use voice input to brainstorm faster and let transcription surface ideas you have already written.
- Plan a coding project in a planning mode before generating code to get more deliberate results.
- Treat AI as a reviewer that critiques your draft rather than as the author of the words.
- Create specialized reviewer personas such as a skeptic or copy editor to examine work from different angles.
- Match tools to strengths: one may be better at structure, another at tone, so combine them.
- Expect English and natural language to become the interface for more of how people use software.
02Key sections
- Voice-first brainstorming
- Petra uses voice input and bilingual transcription to think out loud and surface relevant past writing from her own archive. This keeps ideas flowing and makes her older work a searchable resource.
- Claude Code and planning mode
- Teresa describes how working in the terminal with Claude Code, using planning mode and local files for context, changed her workflow. Planning the project before writing code makes the output more deliberate.
- AI as editor, not ghostwriter
- Both hosts insist they write every word themselves and use AI to review and challenge drafts. The emphasis is on critique and feedback rather than generation.
- Reviewer sub-agents
- Teresa creates sub-agents such as a copy editor, a skeptic, and a devil's advocate to critique drafts from different angles. Each reviewer surfaces a different kind of weakness.
- ChatGPT versus Claude and personal agents
- The hosts contrast ChatGPT's strength in structure with Claude's strength in tone, and discuss personal agents as the future interface for products.
03From the post
“Listen to this episode on: Spotify | Apple Podcasts Teresa Torres and Petra Wille dive into how they’re using AI as strategic thought partners across writing, coding, and product work. Petra shares how she relies on voice input and bilingual transcription to brainstorm and surface her past writing. Teresa explains”
04Frameworks mentioned
Summary and takeaways written by PM Atlas; quotes are short excerpts. © the original author.