How I AI: Teresa Torres’s Claude Code System for Task Management, Automated Research, and ‘Lazy’ Prompting

By: Teresa Torres Host: Claire Vo (ChatPRD) Source: ChatPRD - How I AI Type: article

Summary

Fill after reading. 2-3 sentences: what is this about, what’s the core argument or insight?

Key Ideas Extracted

Fill during processing. Each idea links to a knowledge entry.

Notes

Your annotations, reactions, questions, disagreements. Written during or after reading.

Raw Content

Teresa Torres (author of Continuous Discovery Habits) has built a deeply integrated productivity system using Claude Code + Obsidian. Three core workflows:

Workflow 1: Personalized Task Manager

Custom /today slash command triggers Python script that scans task files and assembles daily to-do list
Every task is its own markdown file with YAML frontmatter (type, due_date, tags)
Tasks created via natural language in Claude Code terminal
Claude handles tagging using taxonomy defined in claude.md
Output: daily markdown file with tasks due today, overdue, in-progress ideas, research digest

Workflow 2: Automated Research Digest

Two Python cron jobs: morning search script (arXiv, Google Scholar) + nightly summarization script
Config file with keywords/topics; tracks what’s already been shown
Claude generates detailed summaries focused on methodology and effect size
New PDFs downloaded manually as a filter step; summaries appear next day in digest

Workflow 3: ‘Lazy Prompting’ via Granular Context Library

Broke context into dozens of tiny, focused markdown files in an “LLM Context” Obsidian vault
Uses index files as maps (e.g., business_profile.md tells Claude where to find specific context)
Global claude.md routes: business question → business profile, personal → personal profile
Built iteratively: at end of each session, asks “Claude, what’d you learn today that we should document?”
Result: simple prompts produce high-quality output because context is pre-loaded and well-organized

Key insight: “pair programming for everything” — Claude Code as a true partner, not just a code tool. The most powerful AI applications may be the ones we build for ourselves.

Referenced tools: Claude Code, Obsidian, VS Code, Descript, Trello Teresa’s links: producttalk.org, justnowpossible.com (podcast), Continuous Discovery Habits (book)