← Back to ClipSAT

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

The one rule everything below exists to protect

"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."

Plus, wherever a figure is load-bearing

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:

50Easy
50Medium
50Hard

Every question in every tier must satisfy all of the following before it merges into the bank:

type

"mcq" or "frq" — multiple-choice or free-response.

domain

The 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 / answer

The question, its options (MCQ only), and the zero-indexed position of the correct one.

sol

The 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.

Status

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.