The AI Skill I Rely On Daily
summary
Analyzes 116 Claude sessions and finds 67% comprehension and only 2% code generation, working in a 15‑year codebase with ~100 PRs merged per day and 100,000 organizations depending on it. Builds a local skill called Catch Me Up—an MD prompt with six exploration modes (architecture, conventions, feature trace, syntax, testing, history)—that returns concrete answers, visual structures, and a demoed loop: verify the agent’s findings before letting it plan and implement.
AI-written summary · about
chapters (10)
- 0:00Introduction and speaker background
- 2:25Sentry's engineering environment and scale
- 3:50AI-driven projects at Sentry
- 5:48Maintaining code quality and technical debt
- 7:35The role of comprehension in software development
- 9:38Analyzing AI usage patterns
- 10:33The "Catch Me Up" skill architecture
- 12:15Short demo of the "Catch Me Up" skill
- 13:56Planning vs. implementation in AI workflows
- 15:26Conclusion and key takeaways
related talks
18mMastering Claude Code for Data Professionals
30mBeyond the Hype: Using AI Coding Agents Today
01h 01m10 tips to level up your ai-assisted coding
01h 00m10 tips to level up your ai-assisted coding
01h 01m10 tips to level up your ai-assisted coding
26mFrom autocomplete to agents: AI coding assistance state of play
25mClaude Skills: Automate the Boring Stuff
32mBeyond the Hype: A Pragmatic Look at How Developers Actually Use AI Tools