I’m one person interested in what open models can do on hardware you can actually own. I work through the difficult parts of running them locally and share anything that might save somebody else time.

What I work on

• MLX and Apple silicon: conversions, quantized checkpoints and architecture compatibility.
• Local inference: practical serving notes, memory observations and fixes tested on real hardware.
• Open sharing: useful artifacts, honest caveats and clear credit to the original model creators.

The local lab

Most of the MLX work happens on a Mac Studio and MacBook Pro. I also use a ThinkStation PGX (DGX Spark) for local inference experiments outside the Apple ecosystem.

If something here saves you a conversion, a compatibility fix or a few hours of testing, then sharing it was worthwhile.

Personal project · Best effort · Upstream licences and attribution always apply