flo_r

5 ms·

1. 1.▲
   Show HN: Clixad – Free AI coding agent in the terminal, funded by an offerwall(clixad.io)

   4 points by flo_r 1mo ago|0 comments
2. 2.▲
   Validating LLM code edits when you can't run the code(velyr.io)

   1 points by flo_r 2mo ago|0 comments
3. 3.▲
   Shipping Shopify theme fixes without write_themes access(velyr.io)

   1 points by flo_r 3mo ago|0 comments
4. 4.▲
   Show HN: Velyr – an AI agent that finds and fixes conversion leaks on your site(velyr.io)

   7 points by flo_r 3mo ago|1 comments
5. 5.▲
   by flo_r 3mo ago

   Gh issues works surprisingly well as an agent board. Labels for state, one issue per feature. The part i haven't figured out yet is how to know when the output is actually done vs just "looks done" to the agent.
6. 6.▲
   by flo_r 3mo ago

   Tried macOS HIG for the same reason and got similar results, less slop, more structure. I think what's happening is that the model has a very specific grammar to pull from instead of averaging over everything web-related it's ever