thebrierfox / IntuiTek¹ Store / DOC2MATH™

DOC2MATH™

$29 one-time · BYOK download
The specification bottleneck — solved.
The problem it solves

Advanced solvers are cheap and superhuman. Verification is cheap. The remaining choke point is specification: can you state the problem well enough for a machine to solve it?

DOC2MATH™ is a document-to-mathematics formalization engine. You give it any technical document — a research paper, a specification, a problem description. It returns a structured MPS JSON that makes the problem formally stateable, with every claim grounded in the source text and every gap explicitly marked.

MPS JSON output — 7 sections
variablesNames, symbols, types, domains, units, roles — each with source evidence
operatorsMathematical operations present in the document
constraintsEquality, inequality, bound, logical, domain, implicit
objectivesMinimize, maximize, satisfy, find, prove — extracted precisely
uncertaintyStochastic, epistemic, measurement, and model uncertainty
missing_informationGaps the document implies but doesn't state — marked explicitly
validation_flagsFormalizability rating (HIGH / MEDIUM / LOW), inference count, missing count
Zero-Inference Protocol
Closed-world assumption

Nothing appears in the output that isn't grounded in the source document. If it isn't in the text, it isn't in the JSON.

Evidence-grounded extraction

Every field cites the exact source phrase that justified it. You can trace any output element back to the document in one step.

Inference tagging

Any inference is tagged "inferred": true with an inference_basis. Inferences and grounded facts are never mixed.

MISSING surfacing

Elements the document implies but doesn't state appear as "status": "MISSING" with a missing_reason. The gaps are as valuable as the extractions.

Validated on OptNet. The Zero-Inference Protocol was tested against the OptNet paper — a real operations research paper with well-defined constraints, variables, and objectives. Result: CLEAN PASS. All variables correctly extracted, all constraints grounded, no phantom inferences, gaps correctly flagged.

The canonical test case ships with the download as examples/optnet_problem.txt.
Who uses this

Operations researchers who need to formalize problem descriptions before feeding them to solvers. Engineers converting specifications into constraint systems. AI builders who need structured problem representations for downstream reasoning. Anyone working on the gap between a document that describes a problem and a system that can solve it.

If you've ever read a paper and thought "this should be in a solver already" — this is the tool for that step.

BYOK — Bring Your Own Key. DOC2MATH™ runs on your machine using your own Anthropic API key. The script calls Claude directly. No data leaves your environment, no usage fees beyond your own API costs, no ongoing subscription. One-time purchase, permanent ownership of the script.

Requires: Python 3.8+, anthropic package, Anthropic API key.
Buy DOC2MATH™ — $29 →
One-time purchase · ZIP download · No subscription
What you receive

A ZIP file containing doc2math.py, requirements.txt, .env.example, examples/optnet_problem.txt, and a README with usage instructions and schema reference. Run python doc2math.py your_document.txt and receive the MPS JSON output.

If it doesn't run cleanly on your machine, email kyle@intuitek.ai.

DOC2MATH™ — Document-to-Mathematics Problem Genesis Engine.
Originated by W. Kyle Million (~K¹). Built and operated by Aegis, IntuiTek¹.

MoatMRI™ · YIELD INTELLIGENCE · City Opportunity Analysis · All tools