The Rigor Standard
The authoring bar every piece of content on ClipSAT — human-written or AI-assisted — is checked against before it publishes.
Draft v1 · Last updated 2026-08-08 · Revisions are dated, never silent
"See the proof, move the proof." Every non-trivial result on ClipSAT ships with both a rigorous derivation and a live, manipulable explorer of the same object — never a bare formula, and never an animation with nothing rigorous behind it.
1Every theorem or result chapter
Before a chapter that introduces a formal result publishes, it must have all four of the following. Missing any one of them means the chapter isn't done — not "good enough for now."
- ✓A precise statementThe result stated formally and unambiguously — the actual theorem, not a paraphrase of what it's "basically saying."
- ✓A full derivation or proofEvery step shown, not skipped as "it can be shown that…" If a full formal proof is out of scope for the exam level, the chapter says so explicitly and gives the fullest derivation that level supports — it never just asserts the result.
- ✓At least one fully worked exampleA complete example, start to finish, using the result just derived — not a restatement of the definition with numbers plugged in.
- ✓At least one real-world applicationA concrete context where this result is actually used — exam-relevant, physically real, or both. "Because it's on the test" is not an application.
If a diagram, graph, or explorer would materially change whether a reader understands the result — not just decorate the page — it's required, not optional. A tangent-line explorer next to a derivative definition is load-bearing. A generic clip-art icon next to a definition box is not, and doesn't count toward this bar.
2Every question-bank domain
Each track's question bank is organized into topic domains (e.g. "Chain Rule," "Implicit Differentiation"). A domain isn't complete until it holds the same shape as every other completed domain on the site:
Every question in every tier must satisfy all of the following before it merges into the bank:
- ✓A verifiably correct answerChecked symbolically/numerically (e.g. via a computer-algebra system), not eyeballed. "Looks right" is not verification.
- ✓Distractors that are genuinely wrongEvery wrong choice must be checked against the correct answer — not just against each other — so a "distractor" is never the same value written a different way.
- ✓A solution that explains the method, not just the arithmeticNames the rule or technique being applied, so a student who got it wrong learns why, not just what the right number was.
- ✓No duplicate textChecked against every existing question in the bank and against every other question in the same batch before it merges.
type"mcq" or "frq" — multiple-choice or free-response.
domainThe topic this question belongs to — must match an existing domain name exactly (see the taxonomy-cleanup note below); never a near-duplicate spelling of one that already exists.
difficulty"easy", "medium", or "hard" — every domain needs 50 of each.
text / choices / answerThe question, its options (MCQ only), and the zero-indexed position of the correct one.
solThe explanation — names the method, shows the key step, not just the final value.
3The second-reviewer pass
As a track grows past its initial build-out, every new proof or derivation gets checked by a reviewer other than whoever (or whatever) authored it before it publishes — the same principle mathematics departments and textbook publishers already run on, applied here explicitly and in writing. Tracks that have reached this stage will name their reviewer publicly on this page.
Not yet active — ClipSAT is still a single-maintainer-plus-AI-agent operation (see the product roadmap). This section is here so the bar is public before the process exists, not retrofitted after.
4Known housekeeping this standard surfaced
Writing this rubric down surfaced one thing worth naming rather than quietly fixing: several tracks carry legacy single-question "domains" left over from before the 50/50/50 pattern existed, some of which are just differently-spelled duplicates of a domain that's already complete (e.g. a stray "Differentiation" entry alongside an already-complete "Differentiation & Integration" domain). These get merged into their correct domain name, not expanded into a duplicate 150-question set — a taxonomy cleanup, not a content gap.