Guadalajara, Jalisco, Mexico. I build local-first agent systems: harnesses, routing, evaluation, and the ops around them. Most of that work stays private. This profile is the public edge of it, not a fork dump.
- Agent harnesses and multi-model seats (ACP, local and hosted).
- Skill routing and evaluation against frozen gold, not vibes.
- Local System One / decision models on a workstation GPU.
I run a lot of experiments. I do not pin other people’s products here.
Original work is private. What you still see under my account as public is mostly forks I used as reading copies. Treat them as that.
Shipping a local System One skill router that has to beat a hosted baseline on a frozen exam, not a demo. If a repo here is empty or a fork, that is the old page. This file is the current one.



