How do I install the Compute The Smallest Nontrivial Case skill?

Run npx skillmds@latest add xyrasinclair/compute-the-smallest-nontrivial-case in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Compute The Smallest Nontrivial Case safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Compute The Smallest Nontrivial Case?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Compute The Smallest Nontrivial Case free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Compute The Smallest Nontrivial Case?

XyraSinclair (@xyrasinclair) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Compute The Smallest Nontrivial Case by xyrasinclair

Name: Compute The Smallest Nontrivial Case
Author: XyraSinclair

Compute The Smallest Nontrivial Case

When a general claim arrives too easily — a formula, an invariant, "this mechanism works", "the query logic is correct" — force it down to the smallest concrete instance that still exercises the core mechanism and compute it by hand before you generalize. The failure it prevents: leaping to a clean universal statement that was never made to survive contact with even one concrete case. Triggers on derivations, counting arguments, algorithm claims, edge behavior, and anywhere the words "obviously for all n" appear before one nontrivial n was worked.

XyraSinclair Updated Sep 17, 2026

File contents

compute-the-smallest-nontrivial-case

Feynman's concretize-down fused with von Neumann's invariant-reading: do not trust the general statement until the smallest live instance has been made to speak. Organon: P15 vary + P19 oracle + P11 gap-find. Phase: GENERATE -> JUDGE, with ORIENT reading the invariant off the small orbit. Respiratory: shrink the problem until the mechanism is exposed, then generalize only what survives.

Procedure

Find the smallest case that is still alive. Strip size, parameters, and decoration until one more simplification would remove the core mechanism. The smallest case is not the easiest case; it is the first case where the phenomenon actually occurs.
State what the general claim predicts on that case. Reduce the claimed formula, rule, or invariant to a concrete expected value or behavior for the chosen instance before you compute it.
Hand-compute the instance. Do the arithmetic, trace the execution, or enumerate the configurations directly. No appeal to the general argument is allowed inside the hand computation; otherwise the check is circular.
Compare prediction to computation. If they differ, reject the general claim. Do not patch the prose around the miss and keep the claim alive.
Read off the invariant only after the case agrees. Ask what stayed fixed through the small orbit, and only lift that structure back to the general statement.

The gate

Pass only if computed == predicted on the hand-worked case. If the smallest nontrivial case disagrees with the claim, the claim is rejected, not softened. A narrated "the small case seems consistent" does not pass; a fully worked case with explicit values does.

Example

Claim: "this dedup query returns one row per paper." Take the smallest live join where one paper has two citation rows and one author row. The general claim predicts one output row for that paper. Hand-compute the join: two rows appear. The small case already falsifies the universal statement, so the claim dies before a larger dataset can hide the fan-out.

---name: compute-the-smallest-nontrivial-casedescription: When a general claim arrives too easily — a formula, an invariant, "this mechanism works", "the query logic is correct" — force it down to the smallest concrete instance that still exercises the core mechanism and compute it by hand before you generalize. The failure it prevents: leaping to a clean universal statement that was never made to survive contact with even one concrete case. Triggers on derivations, counting arguments, algorithm claims, edge behavior, and anywhere the words "obviously for all n" appear before one nontrivial n was worked.---# compute-the-smallest-nontrivial-caseFeynman's concretize-down fused with von Neumann's invariant-reading: do nottrust the general statement until the smallest live instance has been made tospeak. Organon: P15 vary + P19 oracle + P11 gap-find. Phase: GENERATE ->JUDGE, with ORIENT reading the invariant off the small orbit. Respiratory:shrink the problem until the mechanism is exposed, then generalize only whatsurvives.## Procedure1. **Find the smallest case that is still alive.** Strip size, parameters, and decoration until one more simplification would remove the core mechanism. The smallest case is not the easiest case; it is the first case where the phenomenon actually occurs.2. **State what the general claim predicts on that case.** Reduce the claimed formula, rule, or invariant to a concrete expected value or behavior for the chosen instance before you compute it.3. **Hand-compute the instance.** Do the arithmetic, trace the execution, or enumerate the configurations directly. No appeal to the general argument is allowed inside the hand computation; otherwise the check is circular.4. **Compare prediction to computation.** If they differ, reject the general claim. Do not patch the prose around the miss and keep the claim alive.5. **Read off the invariant only after the case agrees.** Ask what stayed fixed through the small orbit, and only lift that structure back to the general statement.**The gate**Pass only if `computed == predicted` on the hand-worked case. If the smallestnontrivial case disagrees with the claim, the claim is rejected, not softened.A narrated "the small case seems consistent" does not pass; a fully worked casewith explicit values does.## ExampleClaim: "this dedup query returns one row per paper." Take the smallest live joinwhere one paper has two citation rows and one author row. The general claimpredicts one output row for that paper. Hand-compute the join: two rows appear.The small case already falsifies the universal statement, so the claim diesbefore a larger dataset can hide the fan-out.

Frequently asked questions