Notes on engineering, architecture, AI-assisted development, and building products.https://erai.dev/https://erai.dev/writing/the-plan-is-the-product/https://erai.dev/writing/the-plan-is-the-product/Most people drive a coding agent by prompting and praying. I run a loop where the code is the last and cheapest step, and a plan document I have argued with is the artifact I actually labor over.Thu, 30 Jul 2026 00:00:00 GMThttps://erai.dev/writing/most-of-your-ai-bill-is-cache/https://erai.dev/writing/most-of-your-ai-bill-is-cache/When your quota vanishes twice as fast as it should, the culprit is almost never how much you generated. It is how often you broke the cache without noticing.Mon, 20 Jul 2026 00:00:00 GMThttps://erai.dev/writing/when-i-reach-for-a-team-of-agents/https://erai.dev/writing/when-i-reach-for-a-team-of-agents/Running five agents in parallel looks like five times the output. Usually it is five times the tokens and a coordination problem. Here is the line I draw.Sat, 11 Jul 2026 00:00:00 GMThttps://erai.dev/writing/agents-need-to-forget/https://erai.dev/writing/agents-need-to-forget/The instinct is to give an AI agent perfect recall of everything you have ever done together. I tried the heavy memory tooling and turned it off. The problem was never storage. It was retrieval.Mon, 08 Jun 2026 00:00:00 GMThttps://erai.dev/writing/the-200-line-rule/https://erai.dev/writing/the-200-line-rule/A long CLAUDE.md feels responsible and reads like diligence. In practice it buries the rules that matter under the ones that do not. Here is the structure I use instead.Mon, 04 May 2026 00:00:00 GMThttps://erai.dev/writing/the-bottleneck-moved/https://erai.dev/writing/the-bottleneck-moved/After a year of building production systems with AI in the loop, the hard part stopped being the code. It became the spec, the context, and the judgment to tell good output from plausible output.Mon, 06 Apr 2026 00:00:00 GMThttps://erai.dev/writing/system-design-in-five-questions/https://erai.dev/writing/system-design-in-five-questions/System design reads like an endless glossary: load balancers, sharding, CAP, circuit breakers, CDNs. It is easier than that. Almost all of it answers one of five questions.Mon, 19 Jan 2026 00:00:00 GMT