DeepMind economist Alex Imas and Ben Golub argue Anthropic’s Fable limits could counterintuitively boost demand for OpenAI models

Also noteworthy: OAI potentially has a playbook - leverage that inference time compute is a knob and make the “xhigh” thinking surpass Fable 5.,.

… now Sol is the orchestrator again.

And given the superior caching/compaction the entire herd could move to OAI - “executor” becomes Terra.

=> subscribers leave Anthropic en masse

…Add Cerebras to the mix, now OpenAI would dominate:

1. Smartest (Sol orchestrator, not Fable)

2. Most cost-effective (Never leave OAI for execution)

3. Fastest

Interesting times ahead

@johnmark_taylor Good point.

But surely some countermoves to that are possible as well?

Eg through pricing; Claude -p is going on a special “credit” budget not the regular token budget

@YanagizawaD The Codex CLI plugin is surely replaceable though. If Claude can type on the terminal it can use Codex

@YanagizawaD OpenAI isn’t losing money on the api lmao

@YanagizawaD I think this would disrupt Codex -> Claude setups but not Claude -> Codex

The deeper point under all this: any setup that bets on one provider's policy is fragile by design. The structural hedge is a multi-provider layer that is also open source, so no single vendor's policy team can strand you. Maestro runs Claude Code, Codex, OpenCode, and Droid side-by-side, and Cue pipelines can route work between the orchestrator and the executor. Give it a try.