4 ms·

Would you dismiss solutions to mathematical problems solved by AI? If I’m driving an AI towards finding a solution, would it be any different for a software pr

Would you dismiss solutions to mathematical problems solved by AI? If I’m driving an AI towards finding a solution, would it be any different for a software project?

spzb 7mo ago
Mathematical proof vs a web app that doesn't actually run? Not much of a contest. Never mind the fact that AIs of the LLM-variety haven't and aren't going to find solutions to mathematical problems.
- ratsimihah 7mo ago
  You got really specific to help prove your point. We were generalising to projects built by AI, not web apps that don’t run, which isn’t relevant since LLMs can clearly build fully working projects. Also how does getting into the specifics of which type of AI can solve mathematical problems helps the comparison here?
  - spzb 7mo ago
    You were the one who made the comparison
- sdoering 7mo ago
  > Never mind the fact that AIs of the LLM-variety haven't and aren't going to find solutions to mathematical problems. This is empirically wrong as of early 2026. Since Christmas 2025, 15 Erdos problems have been moved from "open" to "solved" on erdosproblems.com, 11 of them crediting AI models. Problems #397, #728, and #729 were solved by GPT-5.2 Pro generating original arguments (not literature lookups), formalized in Lean, and verified by Terence Tao himself. Problem #1026 was solved more or less autonomously by Harmonic's Aristotle model in Lean. At IMO 2025, three separate systems (Gemini Deep Think, an OpenAI system, and Aristotle) independently achieved gold-medal performance, solving 5 of 6 problems. DeepSeek-Prover-V2 hits 88.9% on MiniF2F-test. Top models solve 40% of postdoc-level problems on FrontierMath, up from 2%. Tao's own assessment as of March 2026: AI is "ready for primetime" in math and theoretical physics because it "saves more time than it wastes." You can disagree about where this is heading, but "haven't and aren't going to" doesn't survive contact with the data.
  - fredoliveira 7mo ago
    Indeed. And adding on to this, in a slightly different realm, Donald Knuth's conjecture that he solved with Claude: https://www-cs-faculty.stanford.edu/%7Eknuth/papers/claude-cycles.pdf https://www-cs-faculty.stanford.edu/%7Eknuth/papers/claude-c...
  - spzb 7mo ago
    > solved more or less autonomously So, not autonomously.
    - sdoering 7mo ago
      q.e.d.

Would you dismiss solutions to mathematical problems solved by AI? If I’m driving an AI towards finding a solution, would it be any different for a software project?

spzb 7mo ago

Mathematical proof vs a web app that doesn't actually run? Not much of a contest. Never mind the fact that AIs of the LLM-variety haven't and aren't going to find solutions to mathematical problems.

ratsimihah 7mo ago
You got really specific to help prove your point. We were generalising to projects built by AI, not web apps that don’t run, which isn’t relevant since LLMs can clearly build fully working projects. Also how does getting into the specifics of which type of AI can solve mathematical problems helps the comparison here?
- spzb 7mo ago
  You were the one who made the comparison
sdoering 7mo ago
> Never mind the fact that AIs of the LLM-variety haven't and aren't going to find solutions to mathematical problems. This is empirically wrong as of early 2026. Since Christmas 2025, 15 Erdos problems have been moved from "open" to "solved" on erdosproblems.com, 11 of them crediting AI models. Problems #397, #728, and #729 were solved by GPT-5.2 Pro generating original arguments (not literature lookups), formalized in Lean, and verified by Terence Tao himself. Problem #1026 was solved more or less autonomously by Harmonic's Aristotle model in Lean. At IMO 2025, three separate systems (Gemini Deep Think, an OpenAI system, and Aristotle) independently achieved gold-medal performance, solving 5 of 6 problems. DeepSeek-Prover-V2 hits 88.9% on MiniF2F-test. Top models solve 40% of postdoc-level problems on FrontierMath, up from 2%. Tao's own assessment as of March 2026: AI is "ready for primetime" in math and theoretical physics because it "saves more time than it wastes." You can disagree about where this is heading, but "haven't and aren't going to" doesn't survive contact with the data.
- fredoliveira 7mo ago
  Indeed. And adding on to this, in a slightly different realm, Donald Knuth's conjecture that he solved with Claude: https://www-cs-faculty.stanford.edu/%7Eknuth/papers/claude-cycles.pdf https://www-cs-faculty.stanford.edu/%7Eknuth/papers/claude-c...
- spzb 7mo ago
  > solved more or less autonomously So, not autonomously.
  - sdoering 7mo ago
    q.e.d.