Proud to share that Perceptic has been named to The Generalist Future 50 for 2026: 50 of the world's highest-potential startups, nominated by leading investors and selected by @thegeneralistco. We are honored to be in this cohort. 馃殌
1
1
34
AI was supposed to speed up drug discovery. In many R&D teams, it feels like it鈥檚 adding another silo.
1
4
182
That gives the whole team one argument to work on. Biologists, clinicians and commercial leads challenge the claims in their own area and make the call together, and the reasoning stays with the team for the next decision.
1
1
13
These benchmarks test one well-scoped task at a time. They show the agents do that work more reliably inside PercepticOS.
But a real R&D question is many tasks, linked by an argument that has to hold together.
1
1
19
BioMysteryBench, human-solvable: 97.1% vs 89%
BioMysteryBench, human-difficult: 43.8% vs 32%
DrugDiscoveryBench: 80.5% vs 68.7%
1
1
42
Why you shouldn鈥檛 use generalist tools for specialist R&D workflows 馃К
We ran PercepticOS on GPT-6 Astra and compared it with GPT-6 Astra鈥檚 published scores. Same model, so the difference comes from PercepticOS. It scored higher on all three benchmarks 馃У
We just got back from our first offsite in Valencia with the whole Perceptic team.
In January we could all fit around one table. This time there were almost 20 of us, and many more people are starting soon!
We spent two days getting to know each other properly, exploring the city and learning how to cook paella from the locals. Getting a paella right takes the same things we need at work. Everyone owns their part, nobody waits to be told, and you trust the person next to you not to burn the rice.
That鈥檚 the team we鈥檙e building. Mission-oriented, proactive people who want to shape what Perceptic becomes, and put their best work into one of the hardest problems out there, which is getting better medicines to patients faster.
If that sounds like your kind of team, we鈥檙e hiring. Open roles are linked in the comments 馃憞
3
5
364
More AI. More tools. Same silos.
Drug development doesn鈥檛 need another point solution. It needs one place where R&D and commercial teams decide together.
PercepticOS works with what you already have: any model, any cloud, the tools your scientists trust.
Learn more: perceptic.com/os/
Think building blocks for scientific work.
Connect literature, databases, specialist tools and your internal systems through plugins. Add your own tools and models. Build around how your scientists work.
1
3
568
Our science-focused harness scores up to 11.8 percentage points higher than the underlying model鈥檚 published results on BioMysteryBench and DrugDiscoveryBench.










