Introducing BioSecBench-Function, a verifiable benchmark for testing whether AI agents can infer the functional properties of viruses, bacteria, and toxins from data.
Opus 5 with Claude Code leads in endpoint pass rate (50.4%), while Grok 4.6 with Grok Build leads in overall pass rate (44%) when refusals are counted as failures.
The benchmark contains 111 deterministic evaluations spanning viral, bacterial, and toxin systems. It covers five primary threat axes: transmissibility, immune escape, virulence or toxicity, drug resistance, and persistence or fitness.
Each evaluation is grounded in a published study or dataset and reviewed by domain experts. Tasks draw on deep mutational scanning, Tite-Seq, surface plasmon resonance, X-ray crystallography, free-energy calculations, and related assays. Agents receive the relevant data and produce structured answers graded against study-derived ground truth.
Latch turns 3 years old today
In the early days we'd celebrate every time a user ran a workflow
Last year 60+ biotechs ran 30000+ workflows, started 7000+ pods, and spent 50000+ hours on the platform
Far from success, but great memories with the team that makes it possible 🧵
@instagram it is amazing how confusing and imposible it is to get your account back when hacked, my friends buisness account just got hacked and the hacker changed mail and phone number, and it has beensuch a challenge to find a way to communicate with you guys and get the help…
My younger brother, the smarter one in the family, is looking for a SWE internship at a startup this summer.
He published in NeurIPS, is studying CS and Math at UChicago, and has built multiple apps in AI and crypto.
DM me for an intro.
🌎 Hoy la población humana llegó a ocho mil millones.
Leerán a gente "preocupada por la sobrepoblación".
La sobrepoblación NO existe.
Al revés: La humanidad sufre HOY un colapso demográfico.
Y acá te lo demuestro:
1/10
@Dyson pesimo servicio a cliente llevo días intentando comprar un producto y la página no sirve y nadie me contacta para ayudarme, solo mandan reportes y reportes pero nadie me contacta