Fable was the best model I have ever run for zk-autoresearch. Best quality out of ~500 iterations. Never triggered the safety check once on cryptography research. Its first iteration reinvented a known PCS construction from first principles, without reading the papers. It just reasoned its way there.
72 hours later the US government pulled it. Export controls. I'm in Budapest, so if maintained I lose access to the best tool for zk-autoresearch.
This is the AI arms race moving from papers to policy. If export controls hold, the research community splits in half. Open research that was becoming accessible through LLMs slows down overnight.
The alarmism-as-marketing crowd should take a hard look at this. They gave the government the perfect ground for exactly this outcome.