Long-form writing on search, retrieval, agents, and the infrastructure of intelligence.
Strong consistency and conditional writes in blob storage removed much of the need for stateful search clusters in a large class of workloads. They also split search into two problems that need opposite engineering.
I run a small fleet of cloud VMs and a Raspberry Pi at home, wired together with Tailscale. This is about how that setup came together and why Tailscale made it surprisingly simple.
Lately, I have been spending a lot of time with AI coding tools, both cli and ide. I kept asking the question to myself, why is that i keep returning to one tool more than the other. There are lot of blogs around models and their capabilities, but models are also shared across of these surfaces. So the real question for me was that where is the value. Is it in the model or the tooling around it. This blog talks about my explorations and opinions on how to build something competitive.
I have been working on benchmarking lately, especially since I started tinkering with AI. And I came back to an interesting topic of Closed loop vs Open loop benchmarking. To make is easier to...