Can the parallel Agents work on the same problem? I do this in PPLX. I give pplx access to github and have 3 different models (O3, Opus, Sonnet etc) work on the same problem on different branches. Then I let an AI review the PRs and pick the best. Esp for hard problems, this is a life saver.