This matches my experience. Just Talk To It is the best method for working with LLMs if you're an expert. I've seen this at work (as eng manager/lead/principal

This matches my experience. Just Talk To It is the best method for working with LLMs if you're an expert. I've seen this at work (as eng manager/lead/principal/whoevenknowsanymore) – all the big APIs give you stats. We see how much people burn in tokens and we know how much output they produce. There is a pretty strong inverse correlation between token burn and output. The more tokens people burn, the less likely they are to produce a good outcome.