Tech Blog
Local AI experiments, agent systems, and papers worth the time.
Everything here runs on hardware I own. I write up the experiments that worked, the ones that didn’t, and the papers worth reading — with enough detail that you could reproduce them.
What I write about
Local AI
Models running on hardware I own. Quantization, VRAM budgets, what actually fits on a consumer GPU, and the point where local stops being good enough.
Agents
Agent architectures that survive contact with reality: tool design, MCP servers, orchestration, and the failure modes that never make it into the demo.
Papers
Breakdowns of papers worth the time — what the result actually claims, what it quietly does not, and whether it reproduces.
Latest posts
Written by Javier Iracheta — I’ve spent decades testing software, and now I mostly test the machines that write it. More about me →