Originally posted on LinkedIn
One big problem with vibe-coding — it doesn’t maintain patterns and coherence across projects. I think it’s one of the biggest UX opportunities to improve. Slightly technical post 👇
The typical Lovable (or tool X) user workflow: I go in, build what I imagine, iterate over time.
Then, if I want something else, I do the same. And again and again.
But if I want to merge them, it’s a horrible mess. Lovable and similar tools make it very hard (or impossible) to pass another repo so they can “combine” the code — styles, libraries, and patterns from each project can all be different.
So you end up with things like a “central” project linking to 10 “satellite” projects, each with its own look and feel and patterns, each running on its own subdomain (!!!), no shared state (like login!!) between them — total chaos.
With Claude Code (or similar tools) it’s a different story — you can “limit” the context (context engineering, really) to sub-routes very easily and keep everything in the same repo, asking the agent to reference existing styles and patterns for whatever’s being built next.
Lately I’ve wanted to collaborate with non-tech colleagues who are building what they imagine — but the headache of merging the resulting code into the central repo makes me think twice.
How do you handle this? In my case, I’m going to teach my colleagues to “vibe-code” using Claude Code directly on the central repo and have them submit PRs — but it should be easier than this. Don’t Lovable and similar tools have a way to handle this yet?