Don't Become Your Agent's B****
summary
Agents spam noisy pull requests, ask trivial questions, and force constant context switching; model prompts cost a few cents and full runs are about $1, so time is the real cost. Advice is to treat agents like interns: create audited skills and shared utilities, inspect agent trajectories to stop reinventing setup (avoid repeated Gemini 2.5 auth scripts, use Gemini 2.5 vs 3.8 deliberately), prefer fast or hybrid models, duplicate code to silo features, require logs and unit tests, and push agents to run the app and fail-proof PRs.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:33From producer to AI agent user
- 3:59Agents as amnesiac interns
- 6:06Agents slow workflow
- 7:30Taking back control
- 9:56Building agent skills
- 12:04Bad pull requests
- 14:10Code architecture for efficiency
- 16:16Agents fail testing
- 18:21Future of AI agents
- 19:24Seema 2 testing limits
related talks
25mAgents are not Apps: Stop building interfaces, Start building coworkers
40mHow To Tame Your Agents?
34mAI Agents for Performance: Ship Faster, Pay Less
45mAI agents in practice - beyond the prototype
23mAgentic Elixir
30mBeyond the Hype: Using AI Coding Agents Today
39mAgents for the Sake of Happiness
33mYou are the main thread: Multiplying your productivity with agents