Your Attention Is the Bottleneck, Not Your Agents
summary
Claim: human attention limits scale not agents. Uses Claude and Claude Code to read Slack and Linear on a loop, dispatch an agent to an isolated git worktree, run verification gates from lint and unit tests up to browser click‑through and critic passes, and run a weekly agent over JSONL history to surface missing skills. Voice briefs hit 184 words per minute, remote control lets sessions run while away, an Oura ring via MCP feeds sleep data, and the loop demonstrably fixes bugs so you return to completed work.
AI-written summary · about
chapters (16)
- 0:00Introduction and the problem of AI developer burnout
- 1:08A concrete example of using AI agents for bug fixing
- 3:36The bottleneck: human attention vs. infinite agent scalability
- 5:13The proposed stack for sustainable AI development
- 6:03The Signal Layer: Managing notifications and context switching
- 7:04Voice-first flows for coding efficiency
- 8:15The Shower Principle and Remote Control for agents
- 11:13Safety, verification, and testing gates
- 12:05The new, balanced developer workflow
- 13:22System self-improvement using conversation history
- 15:31Holistic well-being (Oura ring integration)
- 17:35Q&A: Addressing skill development early in a career
- 20:07Q&A: Managing JSONL history and long-term conversation logs
- 21:30Q&A: Night shift/background agent execution
- 22:17Q&A: Voice interaction and audio feedback
- 23:39Q&A: Handling complex, multi-stack features
related talks
33mYou are the main thread: Multiplying your productivity with agents
46mWhen You Run Out of Requirements: What Happens When You Go All-In on Agents
23mAgentic Elixir
24mAI Agent Swarms Are Amazing
40mHow To Tame Your Agents?
33mHumans on the loop, not in it: Taking agentic engineering end to end
30mBeyond the Hype: Using AI Coding Agents Today
15mDevelop at Idea Velocity