A lab with a keen question
How should intelligence systems decide what to do before they do it?
How we work
Clarify the ask first
A system should settle what it's being asked, and what is off limits, before it starts generating an answer.
Rules within structure
Limits belong in how a system is built, where input can't rewrite them, not in a prompt that can be overwritten.
Start with people
We borrow from psychology and ethics, how people actually weigh a choice, rather than from next-token prediction on its own.
What we study
- Risk & Security
Keeping intelligent systems hard to subvert, and assessing the misuse and business threats they can create or reveal.
- Intelligence Models
Decision models that borrow from how people reason, not only from what comes next in the sequence.
- Alignment & Governance
Keeping a system inside the limits its user sets, and keeping what it is told apart from what it is shown.
Collaborate
Talk to us
We collaborate with a small number of other deep thinkers, often under NDA.
If your work intersects with ours, we'd love to hear about it.