What do we know?
Claims remain connected to sources rather than disappearing into a summary.
About Cenva
Forward-looking judgment should leave a record.
Cenva exists to make the uncertain parts of research explicit: what a thesis depends on, how likely those dependencies are, what evidence supports them, and what would change the view.
01 / Belief
Research should compound into a better decision model.
Static reports preserve conclusions but lose the structure underneath them. Generic AI can produce more prose, faster. Neither creates a durable record of assumptions, probabilities, evidence, and outcomes.
What do we expect?
Uncertain beliefs become explicit probabilities with visible assumptions.
What actually happened?
Resolved outcomes make it possible to evaluate the quality of the judgment.
Peter Ince
Peter’s work spans AI-driven security, verification systems, and agent architectures. He holds a PhD focused on AI-driven smart-contract security and previously served as CTO at Traverse.
Josh Leavitt
Josh advises Cenva on product and infrastructure strategy, drawing on experience building and operating financial and crypto products at Coinbase.
03 / Operating principles
The product should earn trust through its behavior.
State the uncertainty
Use probabilities, conditions, and caveats instead of false certainty.
Preserve provenance
Keep conclusions connected to the evidence and decisions that produced them.
Expose weak work
Quality problems should change the product state, not be hidden by polished language.
Separate versions
Engine transitions and methodology changes should create clear cohort boundaries.
Keep humans in control
The system structures judgment; it does not own the investment decision.
Publish the misses
A track record only becomes useful when unfavorable outcomes remain visible.
Start with a decision that matters.
We work with investment teams on questions where the assumptions deserve to be explicit.