“That’s why Anthropic fucking lied about Claude Mythos being too powerful to release (it was actually a capacity issue) when in fact it’s just another fucking Large Language Model Nothingburger”
Our reading:
Anthropic lied about Claude Mythos being too powerful to release, when it was actually a capacity issue, and Mythos is just another LLM nothingburger.
Assessment: Contradicted
What happened
Was it true for the period described?
Outcome: Contradicted
The evidence goes against 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
The claim has two parts: that Mythos is 'just another' LLM 'nothingburger,' and that Anthropic lied about its safety reason when the real reason was capacity. On 13 April 2026, the UK AI Security Institute reported its own evaluation (e1). It found Mythos Preview 'a step up over previous frontier models' in cyber skills. Mythos succeeded on 73% of expert-level capture-the-flag tasks (e1). It was the first model to finish AISI's 32-step attack range, which it did in 3 of 10 attempts. It averaged 22 steps, against 16 for Claude Opus 4.6 (e2). These results contradict the 'nothingburger' part. On the capacity part, a 30 April report citing the WSJ said White House officials believed Anthropic had less compute than its competitors (e3). A 13 April analysis called cost and capacity plausible constraints but said they were not shown to be the decisive reason (e5). On 8 May, Anthropic's CEO said compute was not the reason (e4). No item shows that the safety reason was false. The motive part is unproven, and the capability part is contradicted, so the combined claim does not hold.
Caveats
- No item in the packet shows how Anthropic weighed safety against capacity in its release decision.
- AISI says its ranges lack active defenders and defensive tools, so the results may not carry over to well-defended systems.
- The CEO's statement is the company's own account, dated after the post, and is used as context only.
- The compute report is dated after the post and concerns the government blocking wider access.
Evidence (5 items)
Goes against the claim: UK AI Security Institute,
Our results show that Mythos Preview represents a step up over previous frontier models in a landscape where cyber performance was already rapidly improving.
Why it matters: An independent government evaluator found Mythos a step up over previous frontier models. That contradicts the 'nothingburger' claim and shows the cyber concern Anthropic cited had substance.
- Where
- paragraph 1
- Publisher
- UK AI Security Institute
- Kind of source
- Independent measurement
- About events on
- Published
Goes against the claim: UK AI Security Institute,
Claude Mythos Preview is the first model to solve TLO from start to finish, in 3 out of its 10 attempts.
Why it matters: A measured gain over the next-best model (22 vs 16 of 32 steps), and the first model to complete the 32-step attack range. This is a concrete capability advance, not a nothingburger.
- Publisher
- UK AI Security Institute
- Kind of source
- Independent measurement
- About events on
- Published
Context only: The Decoder (citing the Wall Street Journal),
AI advisor David Sacks and other government officials believe Anthropic has less compute capacity than its competitors, according to the report.
Why it matters: Shows that compute was a real concern around Mythos access. This partly backs his capacity point, but it is about the government blocking expansion, not proof that Anthropic's safety reason was false.
- Where
- paragraph 2
- Publisher
- The Decoder (citing the Wall Street Journal)
- Kind of source
- News report
- About events on
- Published
Context only: OfficeChai,
“Mythos consumes a very tiny fraction of Anthropic’s compute. It’s no issue whatsoever to increase the amount of compute by three X or ten X,” Amodei said.
Why it matters: Anthropic's CEO denies that capacity was the reason. This is the company's own self-interested account, so it is context, not decisive.
- Where
- paragraph 3
- Publisher
- OfficeChai
- Kind of source
- News report
- About events on
- Published
Context only: Kingy AI,
Those facts make cost and capacity plausible constraints. They do not show that either was the decisive reason for limiting access.
Why it matters: An analysis that weighs both sides and finds capacity a plausible factor but unproven as the real reason. Context on the 'lied' part.
- Publisher
- Kingy AI
- Kind of source
- Secondary summary
- About events on
- Published
How we judged it
- Kind of claim
- Observation
- Deadline
- None given. We don’t add one.
Where we had to interpret
- The claim has two parts. (1) Anthropic's stated safety reason for restricting Mythos was false, and capacity was the real reason. (2) Mythos showed no meaningful capability advance over other LLMs ('nothingburger').
- Plain reading of 'nothingburger': no notable step up in capability. Independent evaluations of its capabilities test that part.
- The period covers the Mythos Preview announcement on 2026-04-07 through the post date, 2026-04-28. Later statements are context only.
Right if: Credible evidence that Anthropic restricted Mythos mainly for capacity reasons while claiming safety (for example internal documents or credible reporting), and independent evaluations showing Mythos was not meaningfully more capable than existing models.
Wrong if: Independent evaluations showing Mythos was a clear capability step up, especially in the cyber skills Anthropic cited. That undercuts both the 'nothingburger' claim and the idea that the safety reason was invented.
We wrote this test after the outcome was known.
We found 34 claims in this post. See them all.
Full record
Source and dates
Where's Your Ed At, AI's Economics Don't Make Sense, paragraph 179.
- Said
- We read the source
- live page, full text, public
- Evidence up to
- Checked
We read this wording on 2026-10-28. The post may have been edited after 2026-04-28.
Where it was said (1)
Said once in the posts we read.
: Where's Your Ed At, AI's Economics Don't Make Sense
That’s why Anthropic fucking lied about Claude Mythos being too powerful to release (it was actually a capacity issue) when in fact it’s just another fucking Large Language Model Nothingburger
- Where
- paragraph 179
- Published
- 2026-04-28
- We read it
- , live page
- Access
- full text, public
History
Versions
- Current: Contradicted. Evidence up to 2026-10-28.
What happened to this record
PublishedBy us
Published after both AI reviews and the final check passed.