“What was up with multiple graphs (at least two) in which the bars didn’t match the numbers?”
Our reading, the sentence that was checked:
In the GPT-5 announcement, OpenAI showed at least two graphs whose bars did not match the numbers.
Assessment: Supported Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, GPT-5 hot take, paragraph 12.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-gpt-5-hot-take-085a4f-a1
We recorded 8 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 2025-08-07. 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
- multiple graphs
- at least two
- the bars didn’t match the numbers
Our interpretation
- The context identifies OpenAI's August 7 GPT-5 announcement as the event.
- At least two distinct charts must contain bar lengths or heights inconsistent with their displayed numerical labels and scales.
- A mislabeled number can satisfy the claim even if the intended underlying result was correct.
- The claim does not require evidence of deliberate deception or that GPT-5 generated the charts.
Counts as supported if: The original presentation, preserved images, or reliable descriptions identify at least two distinct charts with inconsistent bar geometry and numerical labels.
Counts as contradicted if: A complete original presentation record establishes that fewer than two charts had such inconsistencies.
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
Two distinct announcement charts are documented: the coding-performance chart had a lower labeled value drawn taller, and the coding-deception chart showed o3's 47.4% bar much taller than GPT-5's 50% bar. Together they meet the stated minimum of two. Later corrections do not change what the original presentation displayed.
Limitations
- The assessment relies on contemporaneous visual analysis and subsequent reporting because the primary video was not readable.
- This establishes presentation inconsistencies, not falsified benchmark results, intent, or who made the charts.
Evidence (2 items)
Supports the claim: FlowingData,
52.8% is taller than the one for 69.1%.
Why it matters here: The contemporaneous visual analysis identifies an inverted ordering in the coding-performance chart. This establishes one distinct chart whose bars conflict with its labels.
- Publisher
- FlowingData
- Kind of source
- Independent measurement
- Describes events of
- Public from
Supports the claim: PC Gamer,
roughly three times higher than that of GPT-5.
Why it matters here: The report identifies the separate coding-deception chart: GPT-5 was labeled 50% and o3 47.4%, despite o3's much taller bar. This establishes a second mismatch in the original announcement.
- Publisher
- PC Gamer
- 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, GPT-5 hot take
What was up with multiple graphs (at least two) in which the bars didn’t match the numbers?
- Where
- paragraph 12
- Published
- 2025-08-07
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-gpt-5-hot-take-085a4f-a1: 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.