A numerical answer is a candidate.

Know what
the answer proves.

Check a solution against its exact stored equations. Get a rigorous interval around every component, and see whether your threshold actually follows.

Choose the claim

STAYS IN THIS BROWSER

Start with an original example, or bring a Certloom certificate. The file contains the equations, candidate answer and approximate inverse R.

Loading examples…

Open certificateNo certificate selected

Edit or paste certificate JSON

Finite binary64 coefficients are canonical hexadecimal strings. Duplicate keys, extra fields and arbitrary expressions are refused. Format and limits

Choose an example to begin. No account or runtime download.

Inspect the evidence

The interval decides.

A verified system can still have a wide error bound. A threshold is established only when the entire interval lies on the required side.

CandidateCertified enclosureSeparate decision

Contraction bound · qExact value must be strictly below 1

Transformed residual · ηRecomputed, not trusted

Uniform error radius · rApplies to every component

These decimal quantities are rounded upward. A displayed q of 1 may conceal a smaller exact value; the exact fraction and exact test determine the verdict.

Every component, accounted for

Decimal bounds are rounded outward at 18 places. Expand a row for its authoritative exact fractions. Candidate decimals are for orientation.

Index	Candidate x	Lower bound	Upper bound

Inspect exact proof quantities

Take the evidence with you

The ZIP pairs these exact input bytes with a standalone Python checker, this receipt and instructions. Standard library only; no installation.

Input identity and verification scope

Certificate SHA-256
Meaning: A byte identity, not proof of origin or authenticity.
Checked claim: The exact stored binary64 linear equations, interpreted as rational numbers. This does not certify measurement quality, physical assumptions, an external solver or hardware.

Certify the equations.
Keep the assumptions visible.

The verifier independently recomputes q = ‖I − RA‖∞ and η = ‖R(b − Ax)‖∞ using integers and exact fractions. If the contraction premise fails, it refuses an enclosure. If the interval crosses your threshold, it says undecidable.

Square real systems, 1–128 unknowns. Certificates up to 2 MiB. No uploaded code is executed. A 20-second deadline and Cancel stop only this computation. Serve the source bundle over HTTP; module workers do not run from a file:// page.

Method, boundaries and mathematical foundations ↗