Ask HN: Why aren't we collaborating on the prompts we give to our AI agents?

Now that code is so cheap, I feel like the real work is the spec, the context, the plan (aka the prompt) you hand over to the AI agent. That's where you decide what's actually getting built and how.

That step still seems to be completely individual. You write it alone, the plan you formulate is local and private, and your coworkers only see the result when the PR is opened. On a team, that's the exact moment alignment should happen, and it's the one place we don't work together nowadays.

GitHub Next recently showed a prototype that gets close to this collaborative experience: https://maggieappleton.com/zero-alignment

I also wonder how other knowledge workers work and how they review each others' work. I believe they would also benefit from a "collaborative prompting" environment, where everyone sees what the others are working on and can step in at any point.

Why isn't this normal yet? Missing tooling, cultural habit, or does collaborating on a prompt not help at all?