LocoLLM
Task-specific adapters and a router for small local models.
Open LocoLLM → · Active
Projects
LocoLabo is a set of small, independent projects that share one question: how much useful AI can run on hardware people actually own. Start with the question you have.
- I want to run a model. LocoLLM builds the models, LocoBanco tells you what fits, LocoPuente puts it to work.
- I want to know what old hardware can do. Start with LocoFósil.
- I want to know if it is any good. LocoMusa measures thinking-partner quality.
Build
Capstone
Student teams building domain-specialist adapters for LocoLLM.
Open Capstone → · In progress
LocoEnsayo
AI-powered education simulations and design research.
Open LocoEnsayo → · In progress
Measure
LocoBanco
Which model is best for your VRAM tier, measured on real cards.
Open LocoBanco → · Active
LocoBanco Picker
Plot GPUs by cost, VRAM, bandwidth, power, and benchmark.
Open LocoBanco Picker → · Prototype
LocoFósil
Second-life GPUs, measured in capability, watts, and dollars.
Open LocoFósil → · Active
LocoMusa
Thinking-partner quality for local models, judged pairwise.
Open LocoMusa → · Active
LocoConvoy
Multi-GPU inference on consumer PCIe hardware.
Open LocoConvoy → · Active
Run
LocoAgente
Can small models think in loops? Agent scaffolding research.
Open LocoAgente → · Active
LocoPuente
A local-first AI platform on your own hardware.
Open LocoPuente → · In progress
LocoBase
The Ubuntu setup kit that provisions every lab machine.
Open LocoBase → · Active
Notes
LocoNotas
Field notes from the floor: experiments, reading, and failures.
Open LocoNotas → · Active