A friend who conducts multiple agent swarms recently had his Claude Max account turned off by Anthropic.

He wasn't doing anything against the Terms of Service… and wasn't even using Claude as aggressively as some people I know.

It felt gutting in the moment, but then immediately resolved itself, with a "I'll just use Codex."

A few minutes later, he was back up to speed, as if nothing had happened.

I think of it like the Paul Rudd meme on the computer: "Oh shit! … I'm OK."

Anthropic might think they're buying loyalty with Claude Max subscriptions, but the model is "below the API."

Illustrates that the idea of subsidizing all-you-can-eat usage for a provider's agent harness does not make sense, because no meaningful proprietary state accretes within the agent harness.