Legora CTO hand-writes 100k-param model weights with Fable, one-shot beats training
dair_ai · x · 2026-11-04
Legora CTO Jacsebl demonstrated a striking "weight-space programming" feat: using Fable to hand-write all weights of a 100k-parameter neural network that recognizes real versus fake legal citations — essentially one-shotting a full NN that outperforms training it.
dair-ai's Omar Sansevero, reposting, calls weight-space programming some of the most exciting research he's seen in a long time: instead of training weights, you construct them directly as a readable, programmable object — a novel approach to building small specialized models.
More from Research
- Yann LeCun's Singapore talk breaks down JEPA from the math to robotics applications — ylecun · 2026-11-04
- Solo researcher builds foundation model from scratch using Lie groups and differential geometry — ReportProfessional57 · 2026-11-04
- Nanjing University's SeMo3D Unifies 3D Generation, Editing, and Completion in One Model — nanjinguniv · 2026-11-04
- RoboColosseum Benchmark Finds Weak Cross-Embodiment Agreement in Robot Policies — huggingface · 2026-11-04
- SpecPilot: Berkeley and NVIDIA's Agent Loop Auto-Optimizes Speculative Decoding per Deployment — UCBerkeley · 2026-11-04
- Adobe's MoviEdit Tackles Long-Form Multi-Shot Video Editing With 87K Auto-Generated Pairs — adobe · 2026-11-04