Get the latest tech news
My experience creating software with LLM coding agents – Part 2 (Tips)
My experience creating software with LLM coding agents - Part 2 This post details my experiences creating software with LLM coding agents,...
I cheated and auto-generated the API template code, but now the client and the server have a single source of truth, extremely detailed, with examples of success and failure cases. This forces you to think about architecture ahead of implementation (one of the reasons I hate the term “vibe coding”), but it also addresses one of the AI weaknesses I discussed above- context size. Example (I was trying to fix a “save” bug, but I got worried that the agent was going to make the dialog all reactive and do a crap ton of server calls every time the user typed a character).
Or read this on Hacker News
