CanoProof
The proof-carrying climate decision engine.
For a proposed project area, CanoProof produces a recommendation, the exact evidence behind it, the assumptions it depends on, the smallest change that would reverse it, the claims that must not be made, and a certificate anyone can verify offline without trusting Canopulse.
Self-challenging
The engine attacks its own conclusion: alternative weights, leave-one-component-out, tract exclusion, boundary overlap, aggregation swaps. Survived and failed challenges are both reported.
Reversal frontier
Not a confidence score: the exact smallest change that flips the decision, computed by bisection on the decision itself and reproducible by the verifier.
Independently verifiable
Every analysis exports an evidence package with a standalone verifier. Checksums, frozen-score reproduction, constraints, sensitivity, and reversal all re-check offline.
Run a demonstration analysis
Build a contract, get a certificate.
Selects the highest-need fully scored tracts in the chosen county as a demonstration project area. Inputs are demonstration values; every output is labeled accordingly.
Roadmap discipline
Engines ship only when their evidence exists.
CanoProof is built in phases. Each phase activates only when the evidence it needs is real; nothing is simulated in the meantime.
Active
Decision Proof Engine
Live. Contract, claims, challenges, robustness region, reversal frontier, constraints, minimal repair, certificate, standalone verifier.
Active
Grant Rule Compiler
Demonstration rule library live. Real program encodings require the program document, legal review, and reviewer approval, and are labeled by status.
Active
Intervention Synthesizer
Planning arithmetic with demonstration cost ranges. Any portfolio taken forward must earn its own certificate. No outcome magnitudes are claimed.
Data-gated
Genome and Causal Transport
Data-gated. The Intervention Genome has 0 enrolled projects (enrollment open); causal transport stays unavailable until real before-and-after outcomes exist.
CanoProof outputs are machine-verified under the declared model and constraints. They are screening context: not legal eligibility, not funding guarantees, not causal outcome claims, and not government determinations. Professional review is required before any project decision.