All posts
Everything, chronological. 24 posts.
2026
- The Self-Report Gap: I Audited 4,785 Agent Claims Against What Actually Ran 2026-07-14
- The orchestrator that learns which model to trust 2026-06-23
- Coordinating a fleet of LLM agents on one codebase 2026-06-17
- You can't align what you haven't measured 2026-06-15
- Fable may be smarter. Show me the token bill. 2026-06-11
- The chapter that forgot why it existed 2026-06-08
- Three LLM judges, but really 1.5: why a same-family panel collapses to noise 2026-06-04
- MiniMax-M3: The tier-2 coder that found its niche 2026-06-01
- How we built our user-profile system — the canonical six-layer pattern behind every personalized LLM call 2026-05-27
- We Ran a 3-Source Bug Hunt. Then We Realised Our Validators Were All Claude. 2026-05-26
- Why one agent isn't enough to find your bugs 2026-05-26
- The agent is not a transaction 2026-05-18
- Piaget for prompt agents: why our long-form memory borrows from constructivist psychology 2026-05-16
- Subagents as a context-budget primitive 2026-05-15
- Two prompt frameworks, one runtime: how we adopted BAML without giving up our cost ledger 2026-05-13
- What 170 papers agreed on about deep research agents 2026-05-10
- Mini-ork: A year of autonomous parallel feature delivery on a solo-founder codebase 2026-05-04
- Probe before dispatch: the routing pattern we built without knowing it had a name 2026-05-02
- Our prompt canary was lying to us 2026-04-24
- The paper that proved our 5 lines of code were optimal 2026-04-22
- We stopped treating context like application logic 2026-04-15
- Our prompts stopped being code 2026-01-05
- The simplest survivable form of chat memory 2026-01-04