I’m John Young.
I run engineering at Combine Capital and Infrared, and write research-backed essays on running AI coding agents in production — every claim traced to a primary source.
github.com/johnayoung linkedin.com/in/jyoung1985 john.anto.young@gmail.com rss
Ledger
Essays13
Pillars6
Evidence base verified2026-07-13
Recent Essays
- How to Structure CLAUDE.md: It’s a Loading Policy, Not a Document Agent Runtime Anthropic closed ‘Claude ignores my CLAUDE.md’ as not-planned, area:model. The fix is structural: route every line to the tier that loads it, not a longer file.
- Loop Engineering Breaks Your Single-Shot Context Playbook Agent Runtime A loop was always the agent primitive. The CLAUDE.md budget and JIT retrieval you tuned for one invocation don’t fail louder in a loop — they fail quieter.
- You Can’t Cap What You Can’t Attribute: Per-Task Cost Production Operations A cost dashboard is a lagging report, not a guardrail. Per-task cost attribution is the schema that makes an agent budget ceiling enforceable before the next call.
- Where Just-in-Time Context Retrieval Silently Breaks Agent Runtime JIT context retrieval isn’t free — it’s slower and only as good as your tooling. A ledger of where runtime retrieval breaks and how to design the fallback.
- When One Agent Stops Being Enough: The Isolation Gate Architecture Decisions The trigger for going multi-agent isn’t parallel speed — it’s context isolation, and the split costs about 15x the tokens. Two gates decide whether it’s worth it.