Re:Claude Sonnet 5
by throwaway27448 on Wednesday July 01, 2026 @11:54AM
(#48748892)
I suppose I haven't found much success in complex tasks—it requires so much babysitting, fixing, nitpicking, and modification, I've found it rarely saves much time at all. This is hardly that surprising giving the complexity inherent in describing a task is often explored iteratively, but there's hardly a reason to burn so much money doing so if it doesn't save time.
Perhaps some domains allow a much better tradeoff between complexity of description and execution; perhaps there's a way to get it to exercise better judgement (for lack of a better term) without input from the coder. I just haven't run into this myself.
But they take a lot longer to reach the same goal for complex tasks, so the difference is still very real, and the cost-savings are still very much a question of how well you manage to characterise the tasks they will do quickly and pick and choose what to use when.
I kind of agree that I think the cheap models will eat away at the moat very effectively, but if it doesn't seem more capable to you, you're not giving it complex enough tasks to see what they can do.
(FWIW, I've burned billions of tokens on each of Deepseek, Kimi, GLM5.2, GPT, Sonnet, Opus, Haiku using the same harness, and we've kept stats on cost per task)