The best part of building with AI might be before you write a single line.
I have more fun creating project specs with frameworks like BMAD or GSD than doing the actual coding. I noticed this a while ago and it bugged me until I figured out why.
Writing a spec is pure optimism. Nothing has broken yet. Every feature still fits, the architecture is clean, and the scope is exactly right. You’re describing a version of the project that will never have to exist in reality, which means it can be perfect. That’s genuinely fun.
Then you start building.
The gap between the spec and the thing actually running on your machine can stretch out for a long time. Bugs, edge cases, things that almost work but don’t quite. The fun doesn’t disappear all at once — it drains slowly, which is somehow worse.
I don’t think this means the process is broken. The tedium is just what building things costs. Knowing that doesn’t make the slog shorter, but it at least stops it from feeling like a failure.