jobfit
2 articles on this topic.
- My subagents reminded me of Naruto's shadow clonesI built eight subagents across two repos. In 18 hours they ran 27 times, 11 as real agents: what I delegate, what they found and what I still don't know.
- My prompt couldn't add, and my eval didn't say whyI built an LLM scorer with a threshold I never measured and eyeballed it until something showed up. Two weeks later: 39 labels and a bug in the prompt.