Podcast Summary: How I AI

Episode: Claude Code for Product Managers: Research, Writing, Context Libraries, Custom To-Do System, and More

Guest: Teresa Torres | Host: Claire Vo
Date: January 19, 2026

Episode Overview

In this hands-on episode, product discovery expert Teresa Torres (author of "Continuous Discovery Habits") walks host Claire Vo through her robust, personalized workflows using Claude Code—a generative AI tool—across non-technical and technical realms. Teresa details how she has replaced generic to-do software with her own AI-augmented task management, research aggregation, and writing feedback systems, all powered by Claude, Markdown, and Obsidian. Torres emphasizes practical tips for building “lazy” but powerful systems that supercharge productivity and contextual awareness for product managers and knowledge workers.

Key Discussion Points & Insights

1. The Evolution to Claude Code as an AI Workflow Partner

• Gradual Shift from ChatGPT to Claude

  • Teresa originally used ChatGPT but switched to Claude for improved writing and then integrated it into her technical and non-technical workflows.
  • “I pair program now with everything I do, even if it's not programming. So… I pair task manage and I pair write and I pair everything.” (Teresa, 03:46)

• Integration into VS Code and the Terminal

  • After moving from Trello and basic web GUIs, Teresa now leverages VS Code, Git, and Claude directly in her development environment, streamlining her processes.

• Principle of Pairing with AI

  • “I feel like engineers pair program with Claude when they use Claude code… and I think this idea of pair programming, I pair program now with everything I do…” (Teresa, 03:46)

2. Building a Custom To-Do and Task Management System

• Why Build Your Own?

  • Task management is idiosyncratic; Teresa built a bespoke workflow with Claude to perfectly suit her preferences.
  • “How we manage our tasks is so idiosyncratic that this is exactly the type of thing that you should build for yourself…” (Teresa, 04:41)

• Slash Commands & Daily Routines

  • Created a /today Claude Code shortcut to auto-assemble today’s tasks by scanning Markdown files in Obsidian folders.
  • “Every single morning of my life… I sit down and I literally just type in slash today.” (Teresa, 06:33)

• Markdown + Obsidian for Structure and Access

  • Each task is a Markdown file with a title, date, tags, and context.
  • Obsidian acts as the file browser and knowledge vault; Claude reads and manipulates task files on command.
  • “I kind of think about Obsidian as my file browser. And because it's all in Markdown, it makes everything I do super accessible to Claude.” (Teresa, 11:38)

• Auto-Tagging and Dynamic Views

  • Claude provides tagging for organizational structure and can generate tailored task lists (e.g., sales pipeline) on demand.

Memorable Quote:

“I literally just typed like off the cuff notes to Claude… and because I work in Claude all day, every day, this task window is always open.”
(Teresa, 09:36)

Timestamped Demo Highlights

• Teresa demonstrates creating a task (09:10)
• Obsidian folder structure and use for broader organizational knowledge (11:29)
• Task searching and retrieval powers (15:00)

3. Advanced Research Aggregation—Academic Workflow

• Aspirations as an Academic; Staying on Top of Research

  • Uses Claude-powered daily digests to stay current on topics (synthetic users, team collab, etc.) without manual searching.

• Automated Paper Collection, Summarization & Critique

  • Python scripts (triggered by cron jobs) scrape arXiv and Google Scholar based on custom keywords.
  • Papers are filtered, saved, and then summarized by Claude the next day, with a focus on methodology and effect size.
  • “I get really detailed summaries of… the methods of the paper and the effect size, things that… help me decide, is this worth reading?” (Teresa, 18:30)

• Real Impact Example:

  • Teresa critically reviewed and publicly commented on research, boosting her LinkedIn profile—enabled by her research workflow.
  • “The only reason why I could do that is because I had this system… and it's honestly one of my most best performing posts on LinkedIn ever.” (Teresa, 18:43)

• Potential Expansion to Other Sources

  • Idea to implement similar digests for LinkedIn or market information, hampered by API access issues.

4. Local Knowledge Context—File-Based Memory for AI

• Memory through Context Libraries

  • Torres maintains a structured “LLM Context” Obsidian vault, indexed and scoped by topic (writing, business, persona, product, etc.).
  • AI can load only relevant context files into a session, increasing precision and reducing noise.

• Building and Iterating on Context Files

  • Files created iteratively over time, co-authored by Claude.
  • Contextual scoping prevents overload: “If we give it too much irrelevant context, it’s still going to not be very good at its job.” (Teresa, 31:48)

• “Lazy Prompting” Through Smart Structure

  • Teresa can issue minimal prompts (e.g., “Claude, blog post review”), and Claude fetches the right contextual files to assist.

Memorable Quote:

“I have a Obsidian vault that is literally just for Claude. …The more context I provide to Claude, the more Claude can do for me.”
(Teresa, 25:39)

5. AI as a Writing Partner

• Augmentation Over Automation

  • Teresa enjoys writing but leverages Claude to fact-check, provide critique, enhance hooks, and fix typos.
  • Style guides and writing principles are explicitly codified for personalized, goal-aware feedback.
  • “Claude critiques all of my writing. And by having a really detailed writing style guide, Claude’s critiques are spot on…” (Teresa, 34:24)

• Transparent Use of AI for Some Content

  • In rare cases, large blocks were AI-drafted, but always transparently so.

Lightning Round: Teresa’s Tools, Wishes, and Prompting Tactics

• Favorite AI Toolset:

  • Claude Code (terminal, VS Code), Obsidian
  • Descript for video editing—“just the most magical thing that exists” (38:31)
  • Sometimes ChatGPT in browser out of convenience

• Wish List / API Dreams:

  • Access to LinkedIn’s data via API for research and feed filtering
  • Improved text-to-image generation “where they can close the quotation mark”
  • “I hate AI generated content. I think this is why I still do my own writing, because reading other people's AI generated contents comments kind of breaks my soul…” (Teresa, 39:45)

• Error/Reset Tactics:

  • Uses /clear command to kill context and restart when Claude “gets stuck”—emphasizes the importance of short, evergreen context files for quick resets.

Memorable Quote:

“When Claude gets stuck, I want Claude to go away and I want a clean slate and I want to start over, but I don’t want to have to re-explain all my context to Claude.”
(Teresa, 41:14)

Notable Quotes (with Timestamps)

• “I pair program now with everything I do, even if it's not programming.” (03:46, Teresa)
• “Every single morning of my life… I sit down and I literally just type in slash today.” (06:33, Teresa)
• “I kind of think about Obsidian as my file browser. And because it's all in Markdown, it makes everything I do super accessible to Claude.” (11:38, Teresa)
• “Claude can teach me how to do anything, which I really like.” (40:09, Teresa)
• “When Claude gets stuck, I want Claude to go away and I want a clean slate and I want to start over, but I don’t want to have to re-explain all my context to Claude.” (41:14, Teresa)

Major Timestamps for Reference

• Claude code adoption and workflow philosophy: 02:42–04:14
• Custom slash commands for task management: 06:13–08:51
• File structure & tagging in Obsidian: 09:14–12:56
• Search and retrieval advantages: 14:12–15:29
• Automated research and summarization flow: 16:06–21:44
• Writing with Claude as partner/editor: 32:57–35:14
• Lightning round / daily drivers: 37:11–40:49
• Resetting and context strategies: 41:14–41:49

Conclusion & Where to Find Teresa

Teresa's approach combines lightweight automation, deep context management, and AI augmentation to create a highly productive knowledge and workflow ecosystem, adaptable to any product manager or knowledge worker's unique needs. She blogs at producttalk.org and hosts the podcast Just Now Possible, highlighting cross-functional AI adoption stories.

For actionable takeaways, try building a personal task/knowledge system with Claude or your preferred LLM—start small, iterate, and let the AI do the heavy structuring!