It's apparently lese-Copilot to suggest this these days, but you can find very good hypothesizing and problem solving if you talk conversationally to Claude or

It's apparently lese-Copilot to suggest this these days, but you can find very good hypothesizing and problem solving if you talk conversationally to Claude or probably any of its friends that isn't the terminally personality-collapsed SlopGPT (with or without showing it code, or diagrams); it's actually what they're best at, and often they're even less likely than human interlocutors to just parrot some set phrase at you. It's only when you take the tech out of the area it's good at and start trying to get it to "write code" or even worse "be an agent" that it starts cracking up and emitting garbage; this is only done because companies want to forcememe some kind of product besides "chatbot", whether or not it makes sense. It's a shame because it'll happily and effectively write the docs that don't exist but you wish did for more or less anything. (Writing code examples for docs is not a weak point at all.)