Open is a starting point
Access to weights matters. Access to good interfaces, controls, frameworks, and research matters too.
About SLAI
An independent lab with work to prove.
Skinnertopia Lab for AI researches open models and builds the frameworks and products that help people rely on them.
The beginning
Local AI offered ownership before it offered agency.
Early local models could talk, summarize, and refactor small pieces of code. They could not reliably carry real work. Ollama and Continue helped make AI at home tangible, while the models still left most ambitious tasks out of reach.
Qwen 2.5 showed that open models could begin performing like leading hosted systems. Qwen 3 pushed further through reasoning, tools, and stronger coding. The path toward useful local agents became visible.
SLAI began in August 2025. GRaPE 1 established a series built for ambition. GRaPE 2 caught up with the competition. CRePE extends the work toward code reasoning and project exploration.
The lab exists to keep making better models, and to build the products each model series needs. Open weights should lead to open capability, practical control, and dependable work.
Products are research
A model becomes more legible when it must perform inside a real product with real users and long-running work.
Control belongs to the user
People should decide how deeply a model reasons and how thoroughly an agent acts.
Publish what moves the work
Experiments should leave behind models, methods, evidence, tools, or an honest account of what failed.