Field notes you can run. Everything here will be open source, benchmarked against its own previous version, and published with the failures left in. Take it, fork it, make it your own.
Every tool on this page follows the same three rules. The repo is public and the license is open, no gated tiers, no "contact sales". Every capability claim comes with published benchmark runs, including the iterations that got worse. And nothing here is a product: these are working starting points, built so you can take the idea further than we did.
We're getting the first one ready to ship the right way: public repo, open license, and its benchmark run published alongside it. Follow along on LinkedIn to catch it the day it lands.
This page is meant to fill up with the community's work, not just ours. If you're building an AI-for-defenders tool, or you took one of these somewhere we didn't expect, tell us. That's the whole point.