abhi
99 posts
A bit on how we built it. The mixture: hundreds of thousands of tasks across dozens of environments, weighted toward long-horizon agentic coding in simulated environments that mimic real product interactions and long-running tasks, plus tool use, math, proofs, and broad knowledge. Much of it comes from synthetic environment and reward generation at scale. What we really learned was how to scale data and recipe, and how to deliberately shape model behavior. Intelligence and behavior feed each other in a tight loop a smarter model behaves better, and better behavior unlocks more intelligence. We're just getting started.
We’re sharing the next major milestone in our non-invasive brain-to-text decoder research: Brain2Qwerty v2. Building on v1, which was published today in @Nature, Brain2Qwerty v2 is the highest-performing end-to-end pipeline capable of real-time sentence decoding from raw brain signals. It advances beyond character-level performance to decoding words and semantics, enabling accuracy for overall communication. We believe this research has the potential to make a real difference for the millions of people who suffer from brain lesions or disorders that prevent them from communicating. 🧵👇
We’re sharing the next major milestone in our non-invasive brain-to-text decoder research: Brain2Qwerty v2. Building on v1, which was published today in @Nature, Brain2Qwerty v2 is the highest-performing end-to-end pipeline capable of real-time sentence decoding from raw brain signals. It advances beyond character-level performance to decoding words and semantics, enabling accuracy for overall communication. We believe this research has the potential to make a real difference for the millions of people who suffer from brain lesions or disorders that prevent them from communicating. 🧵👇
