“Alas, “found a solution” didn’t mean what people thought it did.”
Our reading, the sentence that was checked:
Bubeck's claim that GPT-5 'found a solution' to Erdős problems did not mean original solutions as people understood it.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, Erdosgate, paragraph 4.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-erdosgate-201eba-a3
We recorded 6 claims from this post. See every claim from this post.
The wording above is the version this site observed on 2026-10-28. The source may have been edited after 2025-10-21. This page does not claim the wording was captured then.
What the claim commits to
- Kind
- Observation
- Deadline, as the speaker wrote it
- None stated. This site does not add one.
- Deadline state when assessed
- No horizon
- Deadline state on 2026-10-29
- No horizon (the release’s reference date)
From the speaker
- “found a solution” didn’t mean what people thought it did
- People imagined that the system had discovered original solutions to “open problems”
Our interpretation
- People means documented participants or readers, not every reader or a measured majority.
- The relevant distinction is between discovering new mathematical results and locating existing published solutions.
- Assess the announcement and ensuing clarification from the Friday identified in the post through the post date.
Counts as supported if: Records establish that the announced findings were existing solutions and that the wording was understood or publicly presented as new mathematical discoveries.
Counts as contradicted if: The announced findings were original resolutions of previously unsolved problems, or reliable records establish that the alleged misunderstanding did not occur.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Supported
The dated evidence matches what the claim committed to, under the criteria frozen before grading.
Scored on a plain reading. The speaker named no measure for this claim, so it was judged on the reading an informed reader would take, stated below under our interpretation. Both reviews had to find that reading fair.
- Evidence as of
- Deadline state when assessed
- No horizon
- Assessment recorded
- Policy version
- 2026-10-28.7
Rationale
On October 17, 2025, Thomas Bloom said GPT-5 had located references to existing solutions that he had not known about [e1]. On October 18, The Decoder reported that the announcements made it sound as though GPT-5 had independently produced new mathematical proofs [e3]. On October 20, Bubeck acknowledged confusion and explained that the work concerned existing knowledge, rather than AI discovering new results independently [e2]. Together, these records support the distinction Marcus drew between finding existing solutions and the public interpretation of new discoveries [e1, e2, e3].
Limitations
- The evidence does not measure how many readers misunderstood or establish what every reader believed.
- The evidence does not establish intentional deception.
- The report attributes the stronger phrase “previously unsolved” to Kevin Weil; it should not be attributed verbatim to Bubeck [e3].
Evidence (3 items)
Supports the claim: Thomas Bloom, X,
GPT-5 found references
Why it matters here: The database maintainer confirms that the findings were references to existing solutions. His accompanying explanation distinguishes his database's open label from proof that no solution existed.
- Publisher
- Thomas Bloom, X
- Kind of source
- Independent measurement
- Describes events of
- Public from
Supports the claim: Sébastien Bubeck, X,
My posts last week created a lot of unnecessary confusion*
Why it matters here: Bubeck acknowledges confusion and explains that his posts concerned existing knowledge. He also says the earlier nested post explicitly described literature search, which limits any inference of deliberate concealment.
- Publisher
- Sébastien Bubeck, X
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
Supports the claim: The Decoder,
the ambiguous phrase "found solutions."
Why it matters here: The contemporary report explicitly describes how the announcements suggested independently produced proofs. Reporting supplies evidence of public interpretation; the participant statements establish the underlying findings.
- Publisher
- The Decoder
- Kind of source
- News report
- Describes events of
- Public from
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, Erdosgate
Alas, “found a solution” didn’t mean what people thought it did.
- Where
- paragraph 4
- Published
- 2025-10-21
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-erdosgate-201eba-a3: Current. Outcome: Supported. Evidence as of 2026-10-28. Public date not recorded.
Recorded events
Assessment publishedBy this site
First assessment published after both reviews and the final-text check passed.
HeldBy this site
The final-text check failed: citations_correct; overclaims: He also says the earlier nested post explicitly described literature search, which limits any inference of deliberate concealment..
About assessment marcus-erdosgate-201eba-a2, which was not published.