P-01
Local AI workstation
Serving large language models on consumer GPUs, then finding where the software and the models actually break.
Ongoing investigations
Long threads connecting individual experiments, revisions and dead ends.
Serving large language models on consumer GPUs, then finding where the software and the models actually break.
Experiments in accelerating and understanding shell-model calculations without losing sight of numerical correctness.
A long-running workshop in storage, networking, observability, backups and services that have to survive real use.
Field notes will appear here as they are migrated.
Definitions, references and working explanations from computational nuclear physics.
Places and experiences worth recording outside the basement.