Intology's Locus autonomously adapts frontier science AI models to proprietary data
anshulkundaje · x · 2026-11-04
Intology researcher Yusuf Roohani unveiled Locus, an automated research system that autonomously adapts frontier scientific AI models to proprietary datasets, claiming a new frontier in automated molecular design.
Demonstrated use cases:
- Training molecular property prediction models on proprietary tasks
- Fine-tuning Boltz-2 to predict small-molecule potency
- Building antibody design workflows around BoltzGen, Boltz-2, ESMC, and ESMFold2
The team argues that unlocking scientific AI requires leveraging proprietary data, and such tasks are often out-of-distribution for foundation models — Locus handles this adaptation fully autonomously and claims its models outperform SOTA expert solutions.
Related event: Intology Unveils Locus, an Autonomous Scientific Discovery System(14 posts)→
More from Apps
- Grok Bot update lets you shop, pay and track packages inside the chat — XFreeze · 2026-11-04
- CC an AI to book any meeting: how Pepper automates scheduling from email — thechrisperry · 2026-11-04
- Anthropic partners with educator Harper Carroll on series explaining how Claude works — HarperSCarroll · 2026-11-04
- Sonora Studio 1.1 runs YuE2 locally on a 12 GB GPU: free MIT-licensed music studio — baxxox · 2026-11-04
- Apple hosts two Intelligence Labs this week on new Siri AI Developer Preview — rudrank · 2026-11-04
- Dev turns his todo list into an RPG battle where each task is an enemy — chongdashu · 2026-11-04