Activity for jfsiii.bsky.social

Clever!

Embarrassing how long it took to sink in. I refreshed the page a few times 🙃

I can _guess_ what this is offering and ... no, thanks, I'm good

I'm not a planner either, which is what I love about it. I can iterate faster and better than ever before. I can try different shapes, research options, see downstream effects, benchmark choices, compare N options side-by-side, list pros/cons, etc. It tightens the iteration loop for me.

Something that helps during planning are these "theory" docs Still early days but so far so good for humans and bots bsky.app/profile/jfsi...

I mostly do: here's my goal & constraints, let's talk. Ask me questions. Then LOTS of "and what would that look like", "no, like this", etc. Eventually write plan as MD. Start a new session, point at the planning doc and YOLO. Common stuff about needing feedback loops (type check, unit tests, etc)

Interesting. This is not my experience at all. I'm trying to learn about how we can end up having such different interactions. Not just different opinions on the process but seemingly entirely different journeys.

I think some of the AI-related tension I feel is: I care about the code I ship (I've been called a "craftsman", "too precious" about my code) AND I really enjoy using Claude Code. I keep reading it's either/or but it feels like "yes, and" for me. simonwillison.net/guides/agent...