Notes on the AI stack, infrastructure, and the companies building it.
Native Windows support shipped in late 2025, but for anyone whose work involves Docker, bash tooling, or a real Unix filesystem, WSL remains the better path. Here's the walkthrough — what to install, in what order, and the gotchas worth knowing on the way.
Prompting isn't a workaround that goes away as models improve. It's the new literacy layer between business judgment and machine capability — and the productivity gap inside teams using the same AI tools is wider than the gap between teams that use them and teams that don't.
RAG has quietly become the default architecture for enterprise AI. Here's how it works, why retrieval quality is now the real bottleneck, and what to watch as long-context models start to compete with the pattern.