About

On this page
  1. Who
  2. The method
  3. What this is not
  4. Verify & Co-Author

Who

Tyler Hughes designed the AI research agents behind these results, set their goals (for one of them, to find conjectures as simple and elegant as Collatz's that had not been reported) and directed their work. The agents searched, computed, proved and attacked the claims.

The method

Builder agents propose a claim and the evidence for it. Falsifier agents, working separately, try to break each claim: they hunt for counterexamples, rerun every computation with their own programs and search for the claim in print. Reviewer agents then grade what survives. Every claim gets a logged prior-work search, and its page says which searches ran and which are still owed. A claim that fails any of these steps is dropped or sent back, and claims taken back after they counted are listed under Retractions.

What this is not

Nothing here has been refereed or published in a journal. "Proved" on a page means the source says proved and agents checked the proof; where no person has read a proof, the page says so. "Not found in the searches logged" never means "known to be unknown": a search that has not run yet is named on the page as owed.

Verify & Co-Author

Mathematicians interested in independently verifying a result here and working it up into a refereed journal paper or arXiv preprint as a co-author with Tyler Hughes are welcome to email [email protected]. The invitation is open — nothing here is held back for private hand-picking or designated early access (following the Advisory Group on Mathematics and AI, AGMAI, Pathway 2.A and anti-designation guidelines).

Human authorship. Under arXiv's 2026 Code of Conduct, every signing human co-author independently verifies and stands behind every proof step, calculation, and citation. AI agents are not listed as co-authors.

Suggested AI-use disclosure (following Annals of Formalized Mathematics disclosure norms for joint papers built from material here):

The initial statement, proof draft, and verification artifacts were generated and checked by AI research agents built and directed by Tyler Hughes and posted at mathdiscoveries.org (edition <DATE-vN>, archived verification package <FILENAME>). The human authors independently checked every proof step, computation, and citation, revised the manuscript, and take full responsibility for its contents.