Lab
Experiments, in the open.
We run experiments on new models, methods, and tools, and publish every result with the method and data behind it.
Every experiment, on experiments.rubric.shDrag to draw
Findings
Recent results.
- Oct 1467.7%of judge votes went to a small model trained with visual RL
Can a small model learn UI design?
- Sep 2922%fewer tokens with encrypted compaction than text summaries
Which compaction works on long coding tasks?
- Sep 241.43×Jev’s PII masking score, compared with GPT-6 Luna
Can Jev redact PII?
- Sep 3024/24objects passed their schemas, with no text generated
Can Jev do structured outputs?
Newsletter
The Grid
What we’re learning about building with AI. Two short emails a month.