December 16, 2025
Coins, codes, and comment chaos
Erdős Problem #1026
Math legend + AI team-up solves a puzzle, sparks cheers and side-eye
TLDR: Terence Tao shared how a long-unsolved Erdős puzzle was cracked via papers, online collaboration, and AI tools. Commenters raved about the workflow’s real-world value, while purists side-eyed the hype; the loudest theme: AI helps, but human-guided research and fresh sources are essential.
Terence Tao just dropped the saga of how a 1975 Erdős brainteaser got cracked — a coin-piles game where Alice splits the stash and Bob snags “monotone” piles — and the comments lit up. The punchline: it wasn’t AI swooping in alone, but a blend of old papers, online collab, and smart tools like large language models (LLMs, aka text-prediction AIs) guiding fresh research. Read the setup here: Erdős #1026 and Tao’s world: What’s new.
The hottest vibe? “AI-assisted work is the future” — fans cheered that the policy and process are a blueprint for messy real-world problems. One commenter laid it out: static AI memory isn’t enough; you have to re-fetch sources and digest them in context, a shot across the bow of armchair “prompt wizards.” Meanwhile, math purists showed up with polite side-eye: is this proof-by-Google? Not today — the workflow was human-led, with AI as the intern who actually reads the references.
Humor flew: memes of Bob grabbing coin stacks with a “monotone-only diet,” and jokes about “Erdős speedruns” powered by citation bots. The vibe landed somewhere between inspired hackathon energy and gentle turf war: less “AI solved math,” more “the internet — plus AI — did the homework, and the professor brought the receipts.”
Key Points
- •Erdős Problem #1026 was resolved using a mix of existing literature, online collaboration, and AI tools.
- •The original 1975 statement by Erdős was ambiguous, referencing the Erdős–Szekeres theorem and Hanani’s result.
- •The problem appeared on the Erdős problem website on September 12, 2025, with a note on its ambiguity.
- •Desmond Weisenberg proposed a precise formulation using a largest-constant inequality, allowing positive-only and weakly monotone assumptions.
- •A game-theoretic framing equates the problem to maximizing the guaranteed fraction of coins Bob can keep from a monotone subsequence of piles, with discrete and continuous versions aligning in the limit.