Cajal

Cajal is building something the AI world rarely sees: proof, not prediction. Out of YC W26 with co-founder Pedro Nobre, they're applying formal verification — the mathematical method used to prove code is correct beyond doubt — to wrap AI systems running in quantum computing and finance. The bet: the biggest barrier to accelerating scientific discovery isn't better models, it's trust. Instead of hoping an AI is right, you prove it. That's the kind of certainty dense domains have been starving for. We'll see if they can deliver it at scale.