Ventures
I keep coming back to the same move: change the representation and the problem changes with it. Structure that was hidden becomes available to computation, and work that seemed irreducibly difficult becomes mechanical. These three ventures apply that move to healthcare prices, formal mathematics, and post-quantum coordination. A rate keeps its source file, a theorem its assumptions, and an agreement the geometry of the network that produced it.
Palam Health ↗
The data is already public. The problem is that it arrives as terabytes of hospital price files that are technically compliant and practically unusable. Palam turns those files into rate intelligence for the people who negotiate healthcare prices. Every rate stays attached to its source file, version, and timestamp. The first public release is a Colorado procedure search that carries an evidence receipt with every answer.
Guardian Engine ↗
Formal proofs are now produced faster than people can review and organize them. Guardian Engine supplies the missing layer: what a result assumes, whether it already exists, what it strengthens, and how it fits into the library. The whole of Mathlib — 707,053 declarations — has been re-encoded as a content-addressed archive and replayed against the Lean kernel. Results are evaluated by the proof and its assumptions, not by who or what produced them. Mathlib stays upstream; Guardian follows it instead of becoming another fork.
Disentangle ↗
Disentangle derives identity and agreement from the geometry of participation itself. Discrete curvature in the transaction graph makes Sybil clusters visible; topological mass turns genuine integration into consensus weight. That removes the need for mining, staking, or a token, and gives humans and AI agents the same post-quantum identity and delegation substrate. The protocol, SDKs, deployment tooling, and mesh network are implemented in Rust.
