Mathematician Thomas Bloom debunks claim that AI solved Erdős math conjecture

Mathematician Thomas Bloom debunks claim that AI solved Erdős math conjecture · Digg

/TECH

A viral post incorrectly attributed the human proof to "GPT-5.6.

Combined views

2 posts, first seen 20h ago

9 reposts

Commentary on X

Thomas Bloom@thomasfbloom

Which is surprising, yes, but involves no new ideas. Just because a problem is technically 50 years old, and was asked once by Erdos, does not automatically make it significant, or mean that the proof contains anything interesting (above what was already known).

3:19 PM · Jul 14, 2026

Thomas Bloom@thomasfbloom

It's strange what happens to go viral and what doesn't. E.g. I think the other recent GPT/@prz_chojecki proof claim to problem 421 is much more interesting, both in the problem statement and the ideas in the proof (though I haven't verified the latter for correctness yet).

3:23 PM · Jul 14, 2026

idk@DataDon89382

@thomasfbloom the frequency of correct proofs is the most impressive part imo. Went from basically nothing to new news every week

5:23 PM · Jul 14, 2026

Combined views

2 posts, first seen 20h ago

9 reposts

Sentiment

Positive50%50%Negative

Positive Read

Positive users express excitement over GPT-5.6 solving Erdős problem #793 as a sign of frequent correct AI proofs and potential stepping stones, while negative users downplay it as pattern matching without novelty or real advances.

Based on 9 visible X reactions from 13 accounts; directional sample.