Phi9 / independent research & engineering
Understand what
intelligence needs.
What must a system remember to learn and act?
We study representation, computation and control through explicit models, inspectable implementations and experiments that can disagree.
Three connected inquiries
01Information & representation
What must a system remember?
Computation & discovery
When is an answer justified?
Learning & control
How does a policy become useful?
The current evidence
Certainty needs more
than a surviving answer.
A finite discovery study shows why a unique hypothesis can still be wrong. The working manuscript connects that failure to a reproducible offline workflow.
Synthetic, noiseless experiments. Not general discovery, peer review or equal-spend superiority.
Research into practice
A question. A bounded scope.
A result you can inspect.
Bring a representation to evaluate, an experimental assumption to challenge or a research system to implement. We start with the question, the evidence and the permission boundaries.