Blog
Notes from the team
Writing on defensive reinforcement learning, verifiable rewards, and what we're learning as we build.
- Research·September 27, 2026
Isolating the host stops the attack. It also stops the business.
A defender has to stop the attacker and keep the mission running, and scoring only the first trains an agent that pulls the plug. How counterfactual branching produces a reward a machine can check.
ReadPPeter · Co-Founder - Launch·May 29, 2026
Introducing SentryStack
Offense trains on a clean signal and defense does not, so offensive capability is compounding faster. We're building the environments, wargames and verifiers that close the gap.
ReadCConnor · CEO