Anthropic's Mythos arrives, the subsidy era ends, and Anthropic hits $30B ARR

The full Mythos announcement lands with a deliberate non-release. Anthropic raises prices on third-party tool usage. And their ARR triples in just one quarter.

Mythos isn’t getting a public release, and that decision tells you more about where AI is heading than any of the benchmark numbers do.

Anthropic unveils Mythos, but won't release it publicly

My take: Mythos shows massive benchmark gains. 92.1% on Terminal Bench (up from 65.4%). Found and exploited thousands of zero-day vulnerabilities during red-teaming. Anthropic is releasing it only to 40 partners (Apple, Microsoft, Google, AWS, NVIDIA) for defensive cybersecurity. The decision not to ship publicly is the real story. The 'release everything, see what happens' era of AI is ending.

Project Glasswing: a head start for defenders

My take: The frame Anthropic is using is that Mythos's offensive capabilities will eventually exist elsewhere, so giving defensive partners a head start is the responsible move. Whether you buy that or not, the practical implication for cybersecurity vendors is clear. The next 12 months of patching is going to be intense. If you depend on aging software, accelerate your upgrades now.

Anthropic's revenue hits $30B ARR, tripled since end of 2025

My take: That's not just growth, it's growth potentially putting them ahead of OpenAI. Driven almost entirely by enterprise. The take: in any large org, the question 'should we be using Claude?' is no longer a question. The question is which workflows to put it on first.

End of the 'subsidy era'

My take: Anthropic angered some users by charging more for Claude access through third-party tools like Open Claw. They're pushing users to the pay-per-token API. Translation: the era of cheap, all-you-can-eat AI is ending. Build cost monitoring and budgeting into your AI tools from day one. Surprise bills are coming for the unprepared.

OpenAI closes a $122B funding round

My take: $852B valuation. But their secondary market stock is reportedly struggling to find buyers due to the high valuation. Combined with internal disagreement between Altman and the CFO on IPO timing, the OpenAI story is more complicated than the headline number suggests. Worth tracking, especially if you have vendor concentration on OpenAI.

Geopolitical risks for the AI build-out

My take: The war in Iran has declared US tech companies 'legitimate targets,' and US data center projects are being delayed by a shortage of electrical components like transformers and switchgear. The physical supply chain behind AI is more fragile than people realize. Don't assume infinite, instant compute is a given.