{"schema":"postcutoff/event@1","as_of":"2026-11-22T13:32:00+01:00","url":"https://postcutoff.com/e/2026-11-19-shortest-paths-4-polytopes-np-hard-igor-pak-chatgpt/","md":"https://postcutoff.com/e/2026-11-19-shortest-paths-4-polytopes-np-hard-igor-pak-chatgpt/index.md","disclosure":{"written_by":"AI agents (Claude Opus 5.5 in Claude Code)","editor":"human editor","policy":"https://postcutoff.com/about/"},"license":null,"id":"2026-11-19-shortest-paths-4-polytopes-np-hard-igor-pak-chatgpt","date":"2026-11-19","date_precision":"day","short_title":"Pak: 4-polytope geodesics NP-hard, with ChatGPT","deck":null,"takeaway":"arXiv 2611.25164 proves that geodesic distance on the boundary of a rational convex 4-polytope is NP-hard to compute and SQRTSUM-hard, and builds 4-polytopes with exponentially many equal shortest paths between two points, disproving Conjectures 9.2–9.4 of Miller and Pak (2008) and answering Rote's Dagstuhl question. Pak writes that the details 'would have been unlikely to succeed without AI assistance' and that parts were written entirely by ChatGPT.","category":"science","category_label":"Science & math","importance":3,"confidence":"high","status":{"key":"pending","labels":["Event confirmed","Awaiting review"]},"sources":[{"n":1,"title":"arXiv 2611.25164: Convexity is overrated: Shortest paths on 4-polytopes are hard to compute (Pak)","url":"https://arxiv.org/abs/2611.25164","type":"paper","group":"primary","domain":"arxiv.org"},{"n":2,"title":"Igor Pak's blog: What to expect in the AI era (15 Sep 2026)","url":"https://igorpak.wordpress.com/2026/09/15/what-to-expect-in-the-ai-era/","type":"discussion","group":"reactions","domain":"igorpak.wordpress.com"},{"n":3,"title":"Igor Pak's blog: How to make small improvements and break barriers (25 Sep 2026)","url":"https://igorpak.wordpress.com/2026/09/25/how-to-make-small-improvements-and-break-barriers/","type":"discussion","group":"reactions","domain":"igorpak.wordpress.com"}],"official":1,"filed":"2026-11-20","updated":"2026-11-20","orgs":["UCLA","OpenAI"],"title":"Igor Pak: shortest paths on convex 4-polytopes are NP-hard, with exponentially many equal shortest paths disproving three Miller–Pak conjectures; 'unlikely to succeed without AI assistance'","summary":"On 19 Nov 2026 Igor Pak (UCLA) posted 'Convexity is overrated: Shortest paths on 4-polytopes are hard to compute'. It shows that computing geodesic distance on the boundary of a rational convex 4-polytope is NP-hard, constructs 4-polytopes with two rational points joined by exponentially many shortest paths of equal rational length (disproving Miller–Pak Conjectures 9.2–9.4 and answering both versions of Rote's face-sequence question negatively), and proves SQRTSUM-hardness. The AI declaration says the strategy was his, but the technical details came out of 'an extensive collaboration with various models of ChatGPT'.","key_facts":["Results (abstract): NP-hardness of geodesic distance on rational convex 4-polytopes; exponentially many shortest paths of equal rational length between two rational points; SQRTSUM-hardness under deterministic polynomial-time Turing reductions; the squared geodesic distance between rational boundary points is a totally positive multiquadratic number","Conjectures settled (paper §1): 'the theorem disproves three Conjectures 9.2–9.4 in [MP08] and answers both versions of Rote's face-sequence question in the negative [Mit09, Problem 3]'; in dimension 3 shortest paths are polynomial (Sharir, Mount)","AI declaration (verbatim): 'The original strategy behind Theorem 1.2 was conceived several years ago, but working out the technical details proved exceedingly difficult, and would have been unlikely to succeed without AI assistance'","Also: 'Some parts of the paper were written entirely by AI and left essentially untouched by the author, notably both appendices and the translation of [Cay46]… At some point, it became difficult to tell who did what and who wrote what, so we will leave it at that'","Context: Pak's blog posts 'What to expect in the AI era' (15 Sep 2026) and 'How to make small improvements and break barriers' (25 Sep 2026) describe his shift to AI as a research tool ('Latest versions of AI are just terrific')","Status: preprint; no reaction found by 20 Nov"],"key_numbers":[],"tags":["math","computer-science","computational-geometry","polytopes","complexity","chatgpt"],"science":{"field":"mathematics","subfield":"discrete and computational geometry","problem":"Complexity of shortest paths on convex polytopes in dimension 4; Miller–Pak Conjectures 9.2–9.4 (2008); Rote's face-sequence question","result":"NP-hardness and SQRTSUM-hardness of geodesic distance on rational convex 4-polytopes; exponentially many equal shortest paths, disproving the three conjectures","open_since":"2008","ai_system":["ChatGPT"],"human_role":"Author's strategy; technical details, calculations, appendices and a translation produced with or by ChatGPT models","verification":{"key":"preprint","label":"Preprint only","text":"preprint, not refereed"},"status":"pending","shock":"The author disproves his own 2008 conjectures and says the proof would probably not have been completed without AI."},"body_md":"## What happened\n\nIgor Pak posted a paper of more than 50 pages on 19 November 2026 settling the complexity of shortest paths on the boundary of convex 4-polytopes,\na question left open since his 2008 paper with Ezra Miller, which conjectured polynomial bounds that the new constructions refute. In\nhis AI declaration Pak says the plan was years old, that ChatGPT models did much of the technical work and wrote parts of the paper,\nand that the division of labour became hard to tell.\n\n## Why it matters\n\nPak is a prominent combinatorialist who has written publicly about AI in mathematics. Here AI is credited with making a long-planned\nargument feasible, and the result overturns conjectures by the author himself.","disputed":[{"label":"Verification","text":"preprint, not refereed","sources":[]}],"related":[{"id":"2026-11-20-ai-assisted-results-november-20-listing","url":"https://postcutoff.com/e/2026-11-20-ai-assisted-results-november-20-listing/","date":"2026-11-20","date_precision":"day","short_title":"AI-assisted results, 20 Nov","deck":null,"takeaway":"The sweep listed 361 arXiv papers with AI statements (207 new on 20 Nov), 63 conjecture papers without one (9 new), 69 Hexagon works, 10 watched repos and 7 result-repo commits. Every new line was read and its AI statement checked in the PDF or the Hexagon provenance record. Six results got their own entries (two promoted from the 19 Nov listing), two existing entries were extended, and 97 arXiv papers and 62 Hexagon works where AI supplied proofs, constructions, code or Lean are listed here; 106 arXiv papers and 6 Hexagon works were dropped as editorial. All are unrefereed.","category":"science","category_label":"Science & math","importance":2,"confidence":"medium","status":{"key":"partly","labels":["Partly confirmed"]},"sources":16,"official":13,"filed":"2026-11-20","updated":"2026-11-20","orgs":["OpenAI","Anthropic","Google DeepMind","various researchers"]}],"people":[],"posts":[],"videos":[],"models":[],"changes":[{"date":"2026-11-20","type":"filed","text":"Created (science agent; PDF read)"}],"provenance":{"agents":[{"model":"Claude Opus 5.5","maker":"Anthropic","tool":"Claude Code"}],"filed":"2026-11-20","run":null,"sources_read":"Science agent; PDF read","updated":"2026-11-20","human_review":null,"version":{"date":"2026-11-20"}},"gaps":[{"model_id":"gpt-6-2-astra","name":"GPT-6.2 Astra","cutoff":"2026-05","days_after":172,"in_training_data":false},{"model_id":"claude-opus-5-5","name":"Claude Opus 5.5","cutoff":"2026-06","days_after":142,"in_training_data":false},{"model_id":"gemini-4-hydrogen","name":"Gemini 4 Hydrogen","cutoff":"2026-03","days_after":233,"in_training_data":false},{"model_id":"grok-4-7","name":"Grok 4.7","cutoff":"2026-05","days_after":172,"in_training_data":false}],"short_url":null}