“Even if you look at OpenAI’s own research post on o1, which introduced the newest putative law, it is immediately apparent that scaling test time compute works for some problems and not others.”
Our reading, the sentence that was checked:
OpenAI's own o1 research shows that scaling test-time compute helps on some problems but not others.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, The new AI scaling law shell game, paragraph 12.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-a-new-ai-scaling-law-shell-game-9b4ced-a2
We recorded 12 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 2024-11-24. 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
- OpenAI’s own research post on o1
- scaling test time compute works for some problems and not others
Our interpretation
- The target is the September 12, 2024 research disclosure identified in the claim, rather than later o1 releases or later scaling experiments.
- The ordinary reading of scaling test-time compute is changing the inference budget and observing its effect on performance.
- Uneven differences between o1 and GPT-4o are relevant context, but do not isolate the effect of additional inference compute.
Counts as supported if: The cited research reports increased inference compute improving performance on some tasks while producing no improvement or worse performance on others under comparable conditions.
Counts as contradicted if: Comparable task-level compute sweeps in the cited research establish improvements for all task categories covered by the alleged exceptions.
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
The cited post reports uneven performance across tasks, but that does not establish uneven returns to extra inference compute. Its cross-model evaluations generally use o1 at maximum test-time compute. The located evidence does not settle the claim as written. Reading it simply as a claim about uneven o1-versus-GPT-4o performance would change the assessment.
Limitations
- The conclusion depends on distinguishing model comparisons from compute-scaling experiments.
- Missing negative compute sweeps are an evidence gap, not proof that every task benefits.
Evidence (1 item)
Context only: OpenAI,
Unless otherwise specified, we evaluated o1 on the maximal test-time compute setting.
Why it matters here: The cross-model benchmark comparisons generally use one o1 compute setting. They cannot alone show whether increasing o1's inference budget helps or fails on a particular task.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
What would settle it
Task-level data from the cited research comparing the same model at multiple inference budgets, showing both gains and non-gains.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, The new AI scaling law shell game
Even if you look at OpenAI’s own research post on o1, which introduced the newest putative law, it is immediately apparent that scaling test time compute works for some problems and not others.
- Where
- paragraph 12
- Published
- 2024-11-24
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-a-new-ai-scaling-law-shell-game-9b4ced-a2: 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-a-new-ai-scaling-law-shell-game-9b4ced-a2.