HACKER Q&A
📣 consumer451

Why don't LLM dev tools used by "vibe coders" enforce main/develop?


That viral replit story [0] is stuck in my head. Even if we had AGI vibe coding nirvana, we would still need at least a main/develop branching strategy, wouldn't we?

This seems like such an obvious self-own on the part of the tool devs, that I feel like I must be missing some level of complexity in the implementation.

Why do seemingly all of these LLM-based dev tools appear to default to developing in prod?

For a moment I thought that merge conflicts were the reason, but a single dev and simple main/develop git branching strategy, would never have merge conflicts if hot-fixes were not allowed, correct?

Am I missing something, or are these currently caveman dev tools?

[0] https://news.ycombinator.com/item?id=44625119

  👤 threecheese Accepted Answer ✓
My $.02, is these are authoring tools, and not aware of the SDLC. That’s a different/related core competency that the tools would need to perform well in, and because it varies widely it would need to perform well across a broad range of use cases. Git branching is just one small piece, and even that assumes that Git is used and that the user of the tool uses that branching pattern.

Further, this branching pattern would need to be supported across tools - like the MCP server for talking to a database - and now it becomes really complex.

I’m sure we’ll get there, and having these kinds of problems will be what drives businesses to build it.

👤 woodpanel
Since git is very versatile and there are huge differences in written documentation vs practice, as well as practice vs best-practice, I imagine it quite difficult to prompt the LLM in the right direction.

getting git strategies right by an LLM is one of those things of which I am not sure if it is going to be achieved before we have AGI or after that.