“Even setting aside the 1000x costs, not one person outside of OpenAI has evaluated o3’s robustness across different types of problems.”
Our reading, the sentence that was checked:
No one outside OpenAI has evaluated o3's robustness across different problem types.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, o3, AGI, the art of the demo, and what you can expect in 2025, paragraph 24.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-o3-agi-the-art-of-the-demo-and-what-a8b6ba-a1
We recorded 15 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 2024-12-21. 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
- not one person outside of OpenAI
- has evaluated
- o3’s robustness
- across different types of problems
Our interpretation
- This concerns evaluations completed by publication on December 21, 2024, including earlier private evaluations. The beginning of that historical lookback is unspecified.
- In context, robustness across different problem types means testing whether reliable performance extends across meaningfully different domains.
- Merely evaluating different puzzles within ARC does not, on this reading, establish a cross-domain robustness evaluation.
- The claim concerns whether testing occurred, not merely whether a report had been published.
Counts as supported if: Comprehensive access and evaluation records establish that no external person had completed a cross-domain robustness evaluation by publication.
Counts as contradicted if: A documented external evaluation completed by publication tests o3's reliability across meaningfully different problem domains.
Retrospective: these criteria were written after the outcome was public.
Current assessment
Was it true for the period described?
Outcome: Unresolved
The record was checked, and the evidence available by the cutoff does not settle it.
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
External ARC testing is documented, while the announced broader safety-testing program was still beginning. Under the contextual reading of cross-domain robustness, neither fact settles the claim. The absence of a retrieved report cannot prove that nobody had conducted private testing.
Limitations
- A broader reading of problem types could count ARC tasks and produce a different assessment.
- The April source is used only for its retrospective account of December testing.
Evidence (2 items)
Context only: Simon Willison’s Weblog,
soon for full o3
Why it matters here: This contemporaneous account places the announced full-o3 safety-testing program in the future. It cannot exclude earlier private testing under another arrangement.
- Publisher
- Simon Willison’s Weblog
- Kind of source
- News report
- Describes events of
- Public from
Context only: ARC Prize,
tested in December 2024
Why it matters here: External preview testing did occur, and the article describes its ARC scope. This challenges a reading that excludes all external testing, but does not establish testing across broader domains.
- Publisher
- ARC Prize
- Kind of source
- Independent measurement
- Describes events of
- Public from
What would settle it
Dated access and testing records establishing either one qualifying external evaluation before publication or comprehensive coverage showing none occurred.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, o3, AGI, the art of the demo, and what you can expect in 2025
Even setting aside the 1000x costs, not one person outside of OpenAI has evaluated o3’s robustness across different types of problems.
- Where
- paragraph 24
- Published
- 2024-12-21
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-o3-agi-the-art-of-the-demo-and-what-a8b6ba-a1: Current. Outcome: Unresolved. 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.
About assessment marcus-o3-agi-the-art-of-the-demo-and-what-a8b6ba-a1.