About
Lighthouse is one piece a week on building AI systems (and building systems with AI). It’s written for people who build: engineers, PMs, and product folks working with LLMs day to day.
A feel for what you’ll get:
Cutting Claude Costs and Latency with Prompt Caching - the TTL math that decides whether caching saves you 87% or costs you 25% extra
Vibe Coders Have a Reading Problem - on shipping code without reading it
Cosine Similarity Is Dead, Long Live Cosine Similarity - what embedding similarity actually measures
Who’s writing this
I’m Rafael Pierre. I build AI systems for a living, and Lighthouse is where I write down what the work actually looks like: the numbers, the failure modes, and the decisions that don’t make it into the launch post.
Subscribe and you’ll get one issue a week, free. Hit reply on any of them and tell me what you’re building or what’s biting you right now. I read every reply, and it directly shapes what I write next.
