“Coming soon: more on why AI has had less impact on real-world productivity so far than many people had anticipated.”
Our reading, the sentence that was checked:
So far, AI has had less impact on real-world productivity than many people expected.
Assessment: Unresolved Evidence as of
Where and when Gary Marcus wrote it
Marcus on AI, AI’s Black Friday, paragraph 72.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment marcus-ais-black-friday-c7975c-a2
We recorded 11 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 2026-06-06. 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)
Conditions
- so far (qualifier, stated by the speaker)
From the speaker
- AI
- less impact on real-world productivity
- so far
- than many people had anticipated
Our interpretation
- The plain reading is that realized workplace productivity gains had disappointed expectations held by a substantial group. 'Many' need not mean a majority.
- The comparison must concern gains previously expected by the observation period. Expectations for later years cannot establish disappointment by June 6, 2026.
- Productivity means useful output relative to labor or other inputs. Adoption, investment, benchmark performance and employment changes alone do not measure it.
- Read 'so far' cumulatively through the latest complete quarter before publication, ending March 31, 2026. Leave the start unspecified because Marcus supplies none.
- A representative survey comparing earlier expectations with realized gains, or sufficiently broad matched studies, could establish the claim. One small occupational experiment establishes only a narrower example.
Counts as supported if: Evidence shows that a substantial group previously expected greater AI productivity gains by the observation period than actually materialized. Expected and realized gains must concern comparable dates, populations, measures and adoption conditions.
Counts as contradicted if: Broad evidence matching earlier expectations to realized AI productivity gains for the same period shows that expectations were generally met or exceeded and provides evidence against a substantial disappointed group. Positive gains alone do not contradict the claim.
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 evidence establishes a narrow example of disappointed expectations, but does not settle the broader claim. METR's developers expected faster work and instead worked more slowly in early 2025. Its follow-up cautions against carrying that result forward unchanged. The broad firm survey reports small past effects alongside larger future expectations, which is not a missed forecast for the same period. Goldman Sachs explicitly placed the measurable macroeconomic effect later. These records neither establish nor rule out widespread disappointment by the post.
Limitations
- The speaker leaves the relevant population, productivity measure and expected timing unspecified.
- Small experiments and retrospective executive reports cannot by themselves establish a broad expectations shortfall.
- The ongoing observation is assessed through the latest complete quarter before publication.
Evidence (4 items)
Context only: METR,
19% longer
Why it matters here: Sixteen experienced open-source developers expected AI to shorten their work by 24%, but the experiment found longer completion times. This is a matched shortfall in one real workplace setting, not evidence establishing the breadth of Marcus's claim.
- Publisher
- METR
- Kind of source
- Independent measurement
- Describes events of
- Public from
Context only: METR,
some evidence for speedup
Why it matters here: METR judged that developers likely benefited more by early 2026, while emphasizing selection problems and weak evidence about the size of the improvement. This limits extrapolation from e1 but does not establish whether earlier expectations were met.
- Publisher
- METR
- Kind of source
- Independent measurement
- Describes events of
- Public from
Context only: National Bureau of Economic Research,
much larger effects going forward
Why it matters here: This research-publisher summary describes surveys conducted from November 2025 through January 2026: 89% reported no labor-productivity impact over the preceding three years. Its larger expected gains concern the following three years, so that comparison cannot establish missed expectations for the past.
- Publisher
- National Bureau of Economic Research
- Kind of source
- Secondary summary
- Describes events of
- Public from
Context only: Goldman Sachs,
at least 2027
Why it matters here: This prominent forecaster expected the measurable US macroeconomic effect to begin in 2027 and described productivity gains as dependent on widespread adoption over ten years. Modest gains before June 2026 therefore do not demonstrate disappointment against this forecast.
- Publisher
- Goldman Sachs
- Kind of source
- Company disclosure: the company describing itself, not an independent finding
- Describes events of
- Public from
What would settle it
A representative survey or broad set of studies documenting what a substantial group expected AI to deliver by early 2026, compared with realized productivity for the same people or firms, dates, measures and adoption conditions.
Where it was said (1)
Said once in the sources this register covers.
: Marcus on AI, AI’s Black Friday
Coming soon: more on why AI has had less impact on real-world productivity so far than many people had anticipated.
- Where
- paragraph 72
- Published
- 2026-06-06
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment marcus-ais-black-friday-c7975c-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.
HeldBy this site
The openai review found the criteria unfair to the speaker's words.
About assessment marcus-ais-black-friday-c7975c-a1, which was not published.