Source property
Requirement bound to code
Verified quantitative computing
Provable, not plausible.
The verified runtime for AI-generated quantitative code. C Note is the working surface, backed by type checks, independent proof, labeled validation, and evidence a reviewer can inspect.
Gate Type-checks Shape, precision, effects, ownership
Proven
gordon_pv_positive- Domain
- Real arithmetic, not f32
- Sampling
- None
- Contract
- None
Foundation Chelis · Lean-backed language rules
The problem
The review gap
Coding agents produce more quantitative code in a day than a careful reviewer reads in a week. The dangerous errors are silent: a wrong Greek, a flipped hedge sign, or a curve outside its stated bounds. C Proof turns each property into evidence a reviewer can inspect.
The application
C Note
C Note is where a desk writes a quantitative model, runs it, and reviews what was proved, what was validated, and what broke.
The workbench keeps source, requirements, assistant context, and the evidence record in one reviewable artifact.
Evidence
Three explicit outcomes
A deductive result names its proof method. A sampled result stays labeled as validation. A broken property carries a counterexample that can be replayed.
Proven
Proof with its assumptions
Validated
Sampled, not proven
Broken
Counterexample attached
Under the notebook
The verification stack
C Note keeps the working surface close to the machinery that checks each property and records the result.
Surface
C NoteWrite a model, run it, and inspect what was proved, validated, or broken.
Proof engine
Type-checks first, then dispatches each property- Gate · Type system
- Prove · SMT + certified bounds
- Fallback · Labeled validation
Certified envelopes bound supported transcendental subterms; SMT proves the residual. When no proof closes, validation stays labeled.
Foundation
ChelisIts type system rejects structural mistakes; Lean 4 mechanization backs the language rules.
First application
Finance first
The options and derivatives desk prices a wrong number the same day. C Proof checks generated pricing and risk code against the properties the desk says must hold.
- Pricing kernels and Greeks
- Curves, risk measures, and portfolio constraints
- A review record for every stated property
Deployment
Controlled boundary
Controlled deployments keep models, positions, and build paths inside the environment where they already live.
- Runs on infrastructure you control
- No hosted dependency for proprietary models or positions
- No public registry in the build path
Get started
Two ways in
Try C Note
Make an account and prove your first model in the browser. Bring a formula or open a checked example, run it, and see what holds.
Try C NoteTalk to us
Bring a pricing or risk kernel your desk already runs. We will map the properties it has to satisfy and what its C Note shows, on infrastructure you control.
Talk to us