Intent
People rarely know exactly what they want until they see it, and agents fill the gaps with guesses. We study how to surface those guesses while they're still cheap to change.
Fission AI
We're the team behind OpenSpec. We research and build new ways of working with coding agents.
The problems we care about most
Verification
Tests show that code runs. They don't show that it does what someone meant. As agents write more of the world's software, checking it has to get much better.
Judgment
When almost anything can be built quickly, the hard question is what's worth building. We want tools that sharpen that judgment and leave the decision with people.
OpenSpec, our open-source framework, has earned over 70,000 GitHub stars and is used by more than 250,000 developers every month, with a new spec created every two seconds.