Comfortable with the May 2026 rate of AI model progress; he does not call for faster frontier development. [1]
Clarification. A May 2026 interview directly addresses the rate of AI model improvement. It supplies pace evidence absent from the earlier policy essay; this is not evidence of a change in his view.
Source [1] · Podcast interview
Garry Tan
Attributed to: Garry Tan · Tetragrammaton with Rick Rubin
Published
“I don't like to be someone who wants technology to slow down, but right now, at the current rate, I feel good.”
Source context. Discussing whether the AI startup opportunity will last, Tan turns to how quickly AI models get smarter. He says he is comfortable with the rate then, describing it as fast but not so fast that startups or businesses cannot adapt. The show publisher lists this episode on 13 May 2026.
Scope
Rate of AI model capability improvement, discussed in the context of startup opportunities
Conditions
This expresses comfort with the rate as observed in May 2026; it is not an unconditional endorsement of any future rate.
Stated rationale
He says models are improving quickly, but not so quickly that startups or other businesses cannot exist or adapt.
Source details
primary · Original episode as distributed by iHeart. The quoted words and timestamp were cross-checked against a separate automated transcript; punctuation follows that transcript. The audio was not independently transcribed in this review.
Discovered 28 Sep 2026.
Accessed 28 Sep 2026.
Editorial assessment. This interview directly addresses how quickly AI models get smarter. His comfort with the then-current rate and aversion to slowing technology support Continue. His separate advocacy of open development does not establish Accelerate. Saying that progress is slower than some researchers expect is a comparison, not a request to speed it up. The interview does not establish his view for all future rates or all AI policy scopes.
Assessment reviewed
Earlier position ·
INSUFFICIENT EVIDENCE
Supports open AI development and startup competition; no explicit frontier-pace recommendation found in this source. [1]
Initial statement. First recorded observation from the publication dated 2024-06-03. This does not imply an earlier position.
Source [1] · Authored policy essay
Why YC went to DC
Attributed to: Garry Tan · Y Combinator
Published
“It would be far more dangerous for lawmakers to cave to Big Tech lobbying and craft policy that closes off AI development.”
Source context. Support for AI startups or open source is not itself an ACCELERATE or CONTINUE position on frontier model pace. Seek a direct pace statement.
Scope
AI competition, open-source models and startup participation
Conditions
The essay discusses safety concerns but does not prescribe the speed of frontier capability development.
Stated rationale
Tan argues restrictive policy can concentrate power and prevent startup innovation.
Source details
primary · Primary essay authored by Tan; insufficient for a five-category pace classification.
Discovered 26 Sep 2026.
Accessed 26 Sep 2026.
Editorial assessment. Support for AI startups or open source is not itself an ACCELERATE or CONTINUE position on frontier model pace. Seek a direct pace statement.