“I see no signs that the transformer-based architecture can do significantly more than it currently does.”
Our reading, the sentence that was checked:
Zitron says he sees no signs that the transformer-based architecture can do significantly more than it currently does.
Assessment: Unresolved Evidence as of
Where and when Ed Zitron wrote it
Where's Your Ed At, How Does OpenAI Survive?, paragraph 96.
- Said
- Source observed
- live page, full text, public
- Evidence as of
- the cutoff for this assessment
- Assessment recorded
- assessment wyea-to-serve-altman-c971f6-a1
We recorded 38 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 2024-07-29. 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
- I see no signs (qualifier, stated by the speaker)
From the speaker
- I see no signs
- the transformer-based architecture
- can do significantly more than it currently does
Our interpretation
- Assess the public evidence available when the post appeared, not Zitron's private awareness or subsequent AI progress.
- The plain reading is that available research offered no credible indications of substantial further capability gains from transformer-based systems. A sign need not establish that those gains were certain.
- Substantial gains mean materially improved task-solving ability or expanded functionality. No numerical threshold or specific benchmark is supplied.
- The target is the accumulated public research record through July 29, 2024. Its start is unspecified because the claim concerns whether any signs existed by that date.
- Research on training and inference methods counts when it retains a transformer-based model. Results from smaller models are suggestive, but do not automatically establish headroom beyond the strongest contemporary systems.
Counts as supported if: A sufficiently comprehensive contemporaneous research record establishes that apparent routes to substantial further capability gains lacked credible empirical support.
Counts as contradicted if: Research available by July 29, 2024 provides credible empirical indications of substantial capability gains beyond the contemporary baseline while retaining transformer-based models. Prior improvements alone do not suffice.
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
Contemporaneous research offered plausible signs of further progress. Quiet-STaR improved reasoning in a smaller model, and observational scaling research successfully predicted stronger models' capabilities from weaker ones. Those findings weigh against a sweeping absence-of-evidence reading. However, they do not clearly establish substantial headroom beyond the unspecified July 2024 baseline. Research questioning abrupt emergence also does not establish a ceiling. A firm verdict would depend too heavily on supplying a baseline and significance threshold absent from the claim. Later achievements cannot settle this historical observation about the signs available then.
Limitations
- This assesses the public evidence, not what Zitron personally knew.
- The search is not an exhaustive review of all research available before the post.
- Suggestive evidence does not need to guarantee future progress, but its relevance depends on the baseline being assessed.
Evidence (3 items)
Context only: arXiv; Eric Zelikman and colleagues,
Quiet-STaR improves performance
Why it matters here: The paper reports reasoning improvements from training Mistral 7B to generate internal rationales, including an increase from 36.3% to 47.2% on CommonsenseQA. This is a plausible sign of further headroom, but the authors had tested only a seven-billion-parameter model and had not established the benefit at the contemporary frontier.
- Publisher
- arXiv; Eric Zelikman and colleagues
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv; Yangjun Ruan, Chris J. Maddison, and Tatsunori Hashimoto,
agent performance
Why it matters here: Successful predictions across existing models provide an empirical basis for investigating further scaling. The authors explicitly distinguish validation on existing models from predictions awaiting tests on future models; this does not establish substantial gains beyond the July baseline.
- Publisher
- arXiv; Yangjun Ruan, Chris J. Maddison, and Tatsunori Hashimoto
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
Context only: arXiv; Rylan Schaeffer, Brando Miranda, and Sanmi Koyejo,
smooth, continuous, predictable changes
Why it matters here: This study shows how measurement choices can create apparent sudden capability jumps. It supports caution about emergence claims, but smooth improvement remains compatible with substantial gains and does not establish a capability ceiling.
- Publisher
- arXiv; Rylan Schaeffer, Brando Miranda, and Sanmi Koyejo
- Kind of source
- Preprint, not peer reviewed
- Describes events of
- Public from
What would settle it
A contemporaneous statement identifying the intended baseline and meaning of substantial additional capability, together with pre-July 29 research showing or ruling out credible indications of gains beyond that baseline.
Where it was said (1)
Said once in the sources this register covers.
: Where's Your Ed At, How Does OpenAI Survive?
I see no signs that the transformer-based architecture can do significantly more than it currently does.
- Where
- paragraph 96
- Published
- 2024-07-29
- Source observed
- , live page
- Access
- full text, public
History
Assessment versions
- Assessment wyea-to-serve-altman-c971f6-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.