“GPT-5 will not bring us to AGI.”
Our reading:
GPT-5 will not achieve AGI.
Assessment: Supported
What happened
Did the forecast come true?
Outcome: Supported
The evidence matches what they said.
Judged on a plain reading. They named no measure, so we used the reading a typical informed reader would take. It is spelled out under “How we judged it”.
Evidence up to . Checked .
Why
OpenAI released GPT-5 on August 7, 2025 [e1]. That day, The Guardian reported Sam Altman's assessment that GPT-5 had not reached AGI, including his concern about its inability to learn continuously [e2]. METR's July 2025 evaluation, published August 7, tested GPT-5 Thinking before release [e3]. It identified seven tasks with successful human baselines where all six model attempts failed, with no identified technical problems or ambiguities [e3]. These validated capability gaps corroborate the explicit assessment of nonattainment and support the forecast under the contract's reading of AGI [e2, e3]. OpenAI also reported strong software-engineering results, but those domain-specific results do not establish broadly general human-level capability [e4].
Caveats
- The evidence does not supply a universally accepted AGI threshold.
- The explicit assessment of nonattainment comes from OpenAI's CEO, as reported by The Guardian [e2].
- METR tested GPT-5 Thinking on technical tasks; this was not a comprehensive assessment of every GPT-5 configuration [e3].
- METR also found task defects and failures without human baselines; those failures do not carry the same evidential weight [e3].
- The assessment concerns the released system, not future derivatives.
Evidence (4 items)
Supports the claim: The Guardian,
GPT-5 had not reached that goal yet.
Why it matters: Supports nonattainment.
- Publisher
- The Guardian
- Kind of source
- News report
- About events on
- Published
Supports the claim: METR,
There are 7/28 tasks where we are confident that all GPT-5 failures are legitimate.
Why it matters: METR tested GPT-5 Thinking before release and identified seven tasks with human success baselines and genuine repeated failures. This corroborates material capability gaps, although the study was not a comprehensive AGI test.
- Publisher
- METR
- Kind of source
- Independent measurement
- About events on
- Published
Context only: OpenAI,
We are introducing GPT‑5, our best AI system yet.
Why it matters: Establishes that the named system was released and available for assessment. Release alone does not establish whether it achieved AGI.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- About events on
- Published
Context only: OpenAI,
gpt-5-thinking and gpt-5-thinking-mini are our highest-scoring models on this benchmark.
Why it matters: The system card documents substantial software-engineering capability and improvement over earlier OpenAI models. This weighs against portraying GPT-5 as generally incapable, but does not establish AGI.
- Publisher
- OpenAI
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- About events on
- Published
How we judged it
- Kind of claim
- Forecast
- Deadline
- None given. We don’t add one.
Where we had to interpret
- The plain reading is that GPT-5 itself would not achieve broadly general human-level intelligence, including the ability to handle unfamiliar tasks reliably.
- I interpret GPT-5 as the released system bearing that name, including its reasoning capabilities. This does not extend the claim to every successor or future system built using GPT-5.
- Marcus supplied no numerical test or definition of AGI. Strong performance in individual domains does not alone settle whether intelligence is broadly general.
- The forecast has no calendar deadline. Release and substantive capability evaluations can nevertheless settle this model-specific prediction.
Right if: GPT-5 is released, and substantive evaluations identify material gaps in general human-level capability, corroborated by an explicit assessment that the released system has not achieved AGI.
Wrong if: The released GPT-5 demonstrates broadly general human-level intelligence through credible, independently corroborated evaluations across unfamiliar tasks and domains.
We wrote this test after the outcome was known.
We found 12 claims in this post. See them all.
Full record
Source and dates
Marcus on AI, How China might crush the West in the race to AGI if we don’t up our regulatory game, paragraph 4.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-10-29. The post may have been edited after 2023-04-25.
Where it was said (1)
Said once in the posts we read.
: Marcus on AI, How China might crush the West in the race to AGI if we don’t up our regulatory game
GPT-5 will not bring us to AGI.
- Where
- paragraph 4
- Published
- 2023-04-25
- We read it
- , live page
- Access
- full text, public
History
Versions
- Current: Supported. Evidence up to 2026-10-28.
What happened to this record
PublishedBy us
Published after both AI reviews and the final check passed.
Held backBy us
The final check failed: a cited source didn’t say what the write-up said, and the write-up said more than the evidence showed.