The Human Learning Reinforcement Framework (HLRF)
AI should increase human capability without decreasing human understanding. HLRF is our proposed open framework for keeping people's judgment current as AI does more of the work.
Blog
Notes on AI security from the team building it.
Velocity Architecture: A Summary
The promise of AI is velocity. The limit of that velocity is trust. A short tour of our first paper: velocity, trust, limits, decay, and the method that ties them together.
Velocity Engineering: Putting Numbers on Trust
The companion to Velocity Architecture, for the people who build and run AI systems. Once agents are fast, how fast you can safely go depends on how well you catch their mistakes, not on how fast the model is.