Tag
Discussion of dynamic workflows for test-time compute, including their limited use cases, benefits for research experiments, and the need for better benchmarks. Mentions models like Mythos and Opus 4.8 for agent orchestration.