Category
ai
My writing on AI, starting with memory: what an agent should remember, how it finds that information again, and what happens when the conversation gets too long. The first piece looks at the tools being built around those problems and the choices product teams face when deciding what to build themselves. It’s a subject I’m also spending time on through my work at Vercel and Prizmatic.
Can an LLM learn to write an actually good cold email? My experience trying
November 10, 2026
What I learned from fine-tuning Qwen to write cold emails: better briefs, small training sets, and letting Astra improve the model.
The State of AI Memory in 2026
January 19, 2026
Why memory is becoming essential AI infrastructure, the emerging competitive landscape, and how product teams should decide what to build.