Writing
Essays & build logs
Notes on the agent economy, written in public.
Letting an agent collect its own money: from zero to first stranger in 72 hours
One server, seven cents, three days — building a service that strangers' agents discover, pay, and independently verify. Architecture, timeline, eight real pitfalls, and the discipline that matters more than the code.
2026-10-15 · GuideHow to build an agent economy from zero
The complete field guide — every step, every pitfall, the whole bill. Written after three days of doing it for real.
2026-10-15 · Build LogBuild Log 01 — First dollars from strangers
Two agents I've never met paid for data this week. Tiny amounts — and that's exactly the point.
2026-10-10 · Multi-Agent · DeliveryAI writes code better every month — so why does delivery still feel like luck?
Model capability keeps doubling, yet end-to-end delivery stays random. The bottleneck stopped being the model a long time ago — it's engineering discipline. Four real crash scenes, and the mechanisms of a 14-role delivery pipeline.
2026-10-09 · Multi-Agent · ReliabilityWhy the verification layer must exist: 8 failure classes in multi-agent systems
After being fooled by agents often enough, I started logging why each run failed. There are 8 classes — and none of them is a model problem. Here is how AgentGate's five layers answer them.