Locus

Locus is Intology’s automated research system for molecular design, publicly detailed on 2026-11-03. Intology describes it as doing an ML researcher’s loop: propose experiments, run many GPU batches in parallel, evaluate, keep what works, and iterate—especially to exploit proprietary datasets.

Intology claims Locus outperforms specialist stacks (e.g. BioHub/Boltz-class tools) and general LLM coding agents on property prediction, small-molecule potency, and antibody design tasks, framing the product as turning private scientific data into specialized models.

Problem

Drug/molecular R&D often hinges on scarce proprietary data that generic LLMs underuse. Locus aims to automate the model-R&D loop on that data.

Differentiation

Versus one-shot generators or static predictors: closed-loop experiment→evaluate→iterate. Versus general coding agents: science/molecular benchmarks and workflows.

Limits

• Mostly company-reported results.
• Access/deployment via Intology.
• Generic name—disambiguate with Intology when searching.

Who it’s for

Biotech and computational chemistry teams with proprietary molecular data who want an automated research loop.

Get started

1. Read the Intology post and Locus page.
2. Review the @intology announcement thread for claimed benchmarks.
3. Pilot against your specialist baselines on held-out proprietary sets.