Liquid AI's decision recipe matches Jev; new release adds on-device, modalities, primitives
helloiamleonie · x · 2026-11-06
Leonie Monigatti continues the Liquid AI decision-model story: the team now has a recipe competitive with the original Jev model, and in last week's release pursued all three next steps at once — fitting on-device, adding more modalities, and adding more decision primitives. A demo image shows the model answering "what should we do next?" with option probabilities and a calibrated 0.86 confidence score.
More from coding & agent
- Building agent judgement calls with Microsoft Decision-1 scores and thresholds — marlene_zw · 2026-11-06
- Prompt tracking tools may inflate your Search Console impressions via heavy scraping — gaganghotra_ · 2026-11-06
- Android Remote Control MCP v1.13 adds free e2e-encrypted tunnel, no account needed — daniele_dll · 2026-11-06
- MCP auth audit scans 277 servers: 21% accept tools/list with no credentials, 6 criticals found — SarahVoge · 2026-11-06
- Harness-S: rescuing failed agent rollouts via sibling harnesses into one policy — _akhaliq · 2026-11-06
- Meta engineer to speak at DevFest Essex 2026 on AI agents entering Kaggle competitions — AndLukyane · 2026-11-06