We Ran a 10× Bigger Protein Model. It Didn't Rank Variants Any Better.
Scaling ESM-C 600M to 6B gave no zero-shot gain over ESM-2 650M on 201 ProteinGym assays: a non-significant regression (grouped Spearman rho 0.419 vs 0.431).
Read moreVerification-first AI for life sciences.
We build agentic AI systems that prioritize for scientific accuracy.
The product
NeuroAutomata
Score every possible mutation in your protein on ESM-2. Paste a sequence, get a fitness landscape in seconds. Free to start, and your sequences are deleted from our systems within an hour.
Explore NeuroAutomataThe studio
Work with us
We help life sciences and biotech teams move from AI that sounds smart to AI that provides scientific proof: benchmarked accuracy, transparent validation, no black boxes.
See servicesBuilding in public
Read the blogUse NeuroAutomata in Claude and ChatGPT with MCP
Add NeuroAutomata's protein tools to Claude or ChatGPT with one MCP connector URL. Which tools are free, which need Pro, and how the connection is secured.
Read moreAn Independent Check Caught a Claim We'd Overstated
An independent check corrected one of our own live ESM-2 claims about disordered regions — and overturned an internal assessment too. Why we don't self-certify.
Read more