Frontier Models Mozilla shamelessly promotes Mythos. The real story is complicated. Mozilla credits Mythos with 271 Firefox vulnerability fixes. Their own researchers say the harness improved as much as the model. Anthropic conflates the two. Eleanor Hartley · May 8
Frontier Models How OpenAI and others throttle GPU compute and what to do about it OpenAI and others quietly throttle GPU compute behind stable model names. Here's how you can detect it and defend against it (if you're big enough). Eleanor Hartley · May 7
Frontier Models Anthropic's Jack Clark: automated AI R&D likely by 2028 In a striking essay published Sunday, Anthropic's Jack Clark put a 60%+ probability on AI systems training their own successors by the end of 2028. Eleanor Hartley · May 4
