Right now agents are good enough for throwing semi-random ideas at the wall. Experiment compute is the bottleneck because it’s not much more than brute force se

Right now agents are good enough for throwing semi-random ideas at the wall. Experiment compute is the bottleneck because it’s not much more than brute force search. A sufficiently intelligent agent with a deep model of its own architecture will more quickly and confidently locate improvements, the same way that high end LLMs can point out a bug and write a correct fix without even needing to observe and probe the program at runtime. If this level of research performance is reachable, experimentation may become much less of a bottleneck. Hopefully it isn’t.