Exact arithmetic, sealed with a replayable certificate. The model can explain the answer; it cannot override the math.
python -m mathgate "12 * (3 + 4)"12 * (3 + 4)
-> 84 (exact_integer)
sealed sha256:6335247b30de6f51...
No floats, no rounding, no "close enough". 2^64 comes back as a full 20-digit
exact integer, 1/3 + 1/6 comes back as the exact fraction 1/2, sqrt(8)
stays symbolic as 2*sqrt(2) instead of getting rounded to 2.8284271..., and
1 + + * 2 )( comes back REFUSED with a reason instead of a guessed answer.
No dependencies — pure Python standard library.
python -m mathgate "12 * (3 + 4)"
python -m mathgate "2^64"
python -m mathgate --demo
python -m mathgate --json "sqrt(8)"
python -m mathgate --bench # ProofBench-X Lite: 240 seeded, oracle-graded casesEvery result — success or refusal — is sealed with a sha256 hash over the
normalized query, status, and result. Run the same query twice, get the same
hash. That's the whole trust model in one line: the certificate doesn't ask
you to believe the answer, it lets you replay the check.
This repo is a separate, independent implementation covering exact
integer/rational arithmetic, symbolic square roots, integer powers, and
factorial — hand-written recursive-descent parser, graded by an independently
implemented ast-based oracle (mathgate/proofbench_lite.py), 240 seeded
cases, 0 failures, 0 certificate drift, reproducible by anyone.
Those 240 cases are a public development suite: 192 exact answers and 48 expected refusals. The separate oracle cross-checks the tested expressions; this is not an independent third-party evaluation or an official leaderboard result. Targeted regression tests also cover power precedence, nesting, and computation bounds. See the supported limits before using untrusted expressions.
The hosted SuperMath Lab (symbolic calculus, linear algebra, proof mode, and the full 1,116-case ProofBench X corpus) is proprietary and lives at jvi3.com/packages#tab-supermath, with a live scoreboard you can watch update from a real run. See LICENSING.md.
→ Examples · Limitations · 5 bugs a benchmark caught
git clone https://github.com/kyal102/dtl-mathgate.git
cd dtl-mathgate
python -m unittest discover -s tests
python -m mathgate --bench --jsonThat last command is the falsifiable version of "trust us" — run it, and compare the output to what's claimed above and on the hosted scoreboard.
MIT for this lite engine (LICENSE). The hosted engine and its full corpus are proprietary — see LICENSING.md.
