Projects lose hackathon points on things the builder never sees: a README that skips setup steps, claims the code doesn't back up, a first screen that doesn't say what the thing does. Tertius shows you those before you submit. What it does: paste your repo URL, your live URL, and optionally your Devpost text. Judge I reads the repo: files, README and commit history. Judge II loads your live page, measures it and critiques what a reviewer sees first. A claim check then compares what you say with what both judges found. You get a 0 to 100 score (Repo, UI, Base, Claim check, Final) and one ranked fix list. Each fix shows the exact quote or screenshot region behind it.
Why you can trust it: ● Setup, hygiene and commit history are scored by plain code, so the same repo gets the same score every time. ● Every AI quote is checked against the text it was given, and every screenshot region against the image. Anything that fails is dropped. ● Claims nobody could check are marked "Unverifiable" and cost nothing. Tertius never guesses.
