SIDE QUEST 09 / MATH BY COMPUTER / CHECKED SEARCH

Agentic Solvers

Can checking every single case count as real math, if each case comes with a proof a computer can re-check?

The strongest part of this project isn't a long-shot search anymore. It's a package of checked results, ready for expert review. It pushes several published “checked up to this size” limits further in graph labeling. That's a puzzle about giving numbers to the points and lines of a network so they follow certain rules.

STATUSEXPERT REVIEW · RESEARCH
RUNSCERTIFICATES · INDEPENDENT CHECKERS
TRY IT BELOW ↓
01 / SUPER EDGE-MAGIC

Every tree up to 23 points

A tree is a branching network with no loops. The current expert packet confirms that every tree with 2 to 23 points (2 ≤ n ≤ 23) is super edge-magic. That's a special kind of numbering. The project's records say the best published result covered all trees up to 17 points. This is a computer result, checked independently. A specialist still needs to confirm the published record, and a human still needs to review it.

02 / ODD-GRACEFUL

Every tree up to 24 points

The odd-graceful conjecture is a guess that every tree can be numbered in a certain way. The current package confirms it for every tree up to n = 24. The project's check of past work found the best published limit was 12. The main evidence is exact certificate files, plus independent counting and checking. A certificate is a saved numbering for one tree that a program can re-check.

03 / UNICYCLIC GRACEFULNESS

Truszczyński's conjecture up to 19 points

A unicyclic graph is a network with exactly one loop. Every one with at most 19 points is confirmed graceful (another special numbering). The known exceptions are left out. Those are single rings Cn where n leaves a remainder of 1 or 2 when divided by 4 (n ≡ 1 or 2 mod 4). The project records n ≤ 9 as the earlier full-check limit it could confirm. That still needs the same expert check of past work.

04 / REPRODUCIBILITY

Independent checkers, not one solver

The certificate files can be rebuilt exactly, every time. A separately written checker, in the Rust language, checks each numbering. It confirms no two trees are the same shape, using a standard ID for each shape (a canonical key). It also checks the total against Otter's known tree counts. The unicyclic result has two separate checkers, one in Python and one in Rust.

MATURITY / LEVEL 11

Ready for experts isn't the same as peer reviewed.

My wider publishing project rates this tree package level 11: ready for expert review. That means the AI-run checks, rebuilds and handoff prep are done. So are the red-team attacks, where AI agents try hard to break the result. What's left needs people: searching specialist paper databases, expert review, deciding on authorship and how to disclose AI use, and picking where to publish.

WHY THIS IS INTERESTING

It pushes a checked limit further.

I'm not claiming a proof for every size. The result is narrower and easier to audit. Every case in clearly stated size ranges was checked. Each graph has its own certificate, and separate checks confirm that no graph was missed.

RESEARCH SYSTEM

The solver is only half the experiment.

Finding and approving are kept apart. One program may suggest a numbering, but a different program checks it. Rebuild tests make sure nothing depends on one single compiled program. Red teams and negative controls go looking for ways it could fail. In a negative control, you already know the right answer is no. Last, a review of past work decides whether the extended range is actually new.

CURRENT HANDOFF

Three results, each checked up to a set size, are waiting for experts.

The package is frozen for expert review. AI agents don't keep rewriting it, and that's on purpose. Once the automatic checks are done, changes should only come from a reported mistake, newly found past work, or human feedback.

$ agentic-solvers --statussuper edge-magic ... trees n ≤ 23odd-graceful ...... trees n ≤ 24unicyclic graceful . n ≤ 19, stated exceptionsindependent checks . PASSlabel .............. EXPERT_REVIEW_READYpeer review ......... pending
← Agentic MathsAll Side QuestsSLAY2 →

The Inner Universe

Still in the works.

I’m still writing this part of the site: a map of the big questions I keep coming back to. Check back soon.