If you haven't read "The Mom Test," stop building and read it this weekend.
It's about how to talk to users without them lying to you to be nice. Single biggest fix to how I do customer interviews.
For anyone building with LLMs — the prompt engineering docs from Anthropic are underrated.
Spent a weekend on them and cut my token usage ~40% just by structuring prompts properly. Free, no signup.
First paying customer found me through a reddit comment I almost didn't post.
Lesson: show up where your users already are, even when it feels like shouting into the void.
TIL Claude has a "Projects" feature that keeps context across chats.
I'd been re-pasting the same background into every conversation like an animal. Dropped my project docs in once, now every chat already knows my stack and constraints. Should've found this months ago.