Lou Rossi

Claude Code is amazing. Agentic workflows just... work - at least better than last years options for vibe coding.
Things I've learned:
- .md files are a must
- architecture and context
- log history after every pass
- Short passes with initial creation, full pass with design is fine
- Fable is a hog.
- use ChatGPT or Gemini for UI/UX and design prompting, but not before you get Claude to chop down your architecture first to hand off.
- /clear
What other things do you use to ensure Claude is productive and doesn't break with a long term project?

My terminal coding agent had hit its weekly limit and I had a small bug sitting in my queue, the kind of thing that would normally take one prompt and thirty seconds, so I figured I'd just fix it myself while I waited.

Took me almost the full thirty minutes. Not because the bug was hard, it genuinely wasn't, but because I kept reaching for a keyboard shortcut that wasn't there and realized I'd forgotten the actual syntax for something I used to type without thinking. Had to look up documentation I would've known cold two years ago.

Got it fixed right as the coding agent's limit was reset and honestly felt a little relieved, like I'd passed some test I didn't know I was taking. But it also left me wondering how much of that muscle is still in there versus how much I've quietly outsourced without noticing, since it's not like there was ever a single moment where I decided to stop practicing it.

Has anyone else tried this on purpose, going a day without the assistant just to check? Would you actually want to know the answer, or is it better not to look too closely at how much you've forgotten?

Because building products has become much more accessible thanks to AI, the bar has had to rise a bit, or at least the expectations have become stricter.

I built my own tool, but personally, I wouldn t launch it on Product Hunt yet because it s still very basic and doesn t even have a website.