The use of AI Agents creates a distinctive smell... One can tell the GH Repo owner was high on Claude just by looking at verbose and hard to follow README.md lacking clarity and brevity. My weekend experiment cutting 40% of lines of code (without compromising the functionality) from an AI grown codebase is an eye opening experience into what AI bloat might look like. The learnings have been distilled into an agent skill.
Last autumn I started building Flutter app entirely with AI - a media player. I would not say I vibe-coded it - I pressed agents to keep up docs, pushed automated tests coverage, invested in feedback loops (e.g. created ergonomic CLI for Flutter app driving).The thing could be run and poked from the outside. There was structure around the agents.
But I also did not read the code very much - I was too lazy. Or, more precisely, reading the code felt like opening a portal. Once you start looking, you do not just "review" it. You notice weird layers, half-fixes, old ideas still wired through the system, comments explaining nothing, abstractions introduced for a problem that no longer exists, and then the choice becomes: do I stop and rewrite this? Do I spend the weekend paying down debt I only discovered because I looked? So I kept shipping around it.
