Post
What #LLM development looks like for me:
⁃ describe a pair of feature changes I want to make, with code and GitHub issues provided
⁃ let the LLM churn on it for a bit, iterating with a live copy of the site in front of me
⁃ after about 2 days of iteration, including carefully reviewing the outputted UI and docs, get to the point where I can review the PR code
⁃ 2 more days of carefully reading the PR, cutting out redundant code
It's not magic, but with care it can serve a purpose.
I truly don’t understand vibe coders; I can’t bear the idea of letting the first pass junk that I got out of this stand as something with my name on it.
Did it get me a good start? Yes. Did I need to revise everything from the design and docs through the naming and implementation? Also yes.
@offby1 My experience exactly. I can get Claude to dump out 90% of the code I need for a feature, but then it takes me substantial time to actually review it and beat it into shape as something presentable to another human for review. It's a huge time saver only if you don't care what you commit, and that's not me.
That said, I do see a benefit in my focus management. I find the AI makes a lot of small decisions in that initial 90%, and many of them aren't important. That saves mental energy.
This is a bonfire demo instance for testing purposes. This is not a production site. There are no backups for now. Data, including profiles may be wiped without notice. No service or other guarantees expressed or implied.
