# Intology's Locus automates molecular model design and beats baselines and LLM agents

> Original title: An AI system just designed better molecular models than the specialist tools built by the labs that own those models.

- Source: Rohan Paul
- Published: 2026-11-03T17:05:23.000Z
- Bad Influence: https://badinfluence.news/items/o1tpgepras2u7aj9uge2jo5ku
- Original: https://x.com/rohanpaul_ai/status/2117664448609665321

## Summary

Rohan Paul reports that Intology's automated research system Locus outperformed state-of-the-art baselines and LLM agents on molecular design tasks. The tasks include property prediction, fine-tuning Boltz-2 for small-molecule potency, and building antibody design workflows. Locus is described as running experiments in parallel on GPUs and keeping what works, and the post includes a benchmark figure comparing it with Codex, Claude Code, Boltz-2 API, BoltzGen, and ESMFold2.

## Post

An AI system just designed better molecular models than the specialist tools built by the labs that own those models.

@intology’s Locus does a machine-learning researcher’s job: 
> proposes experiments, 
> runs many batches of them in parallel on GPUs
> keeps what works and repeats, 
> now applied to drug and antibody design.

Locus beat GPT-6 Astra and Claude Opus 5.5 agents on molecular prediction

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWNzEuObUAASzX-.jpg&mode=thumb&exp=1793923200&sig=3ab3f27c6444b5ec)

![](https://badinfluence.news/api/img-proxy?u=https%3A%2F%2Fpbs.twimg.com%2Fmedia%2FHWNzJr9bUAIYEBz.jpg&mode=thumb&exp=1793923200&sig=01af90b639572da2)

## Quoted post @intology

> Locus sets the frontier for automating molecular design
> 
> Our automated research system Locus develops scientific AI models that outperform SOTA expert solutions
> 
> Unlocking the potential of scientific AI requires leveraging proprietary data. Locus achieves this autonomously 🧵👇
> 
> --
> 
> We test Locus across a number of tasks related to molecular scoring and design: training molecular property prediction models for a proprietary task, fine-tuning Boltz-2 to predict small-molecule potency, and building antibody design workflows around BoltzGen, Boltz-2, ESMC, and ESMFold2.
> 
> Each of these settings are real-world AI for science use cases, where tasks are often out-of-distribution for existing foundation models and limited in data, while featuring expensive and noisy evaluators. Compared to state-of-the-art baselines and LLM agents set to optimize these tasks, Locus achieves higher performance. Additionally, Locus generalizes better despite limited data, finds more sample efficient solutions, and explores a higher diversity of solutions.

https://x.com/intology/status/2117642701890105725