We’re not writing code anymore. We’re directing it.

The best mental model I've found for AI-assisted development: you're a film director. AI Coding is Directing.

You have a vision in your head. Your job is articulating it clearly enough that your crew (the AI) can execute. The clearer your direction, the better the output. Vague instructions get vague results. Specific prompts get specific implementations.

What Changed

The models got good enough that directing beats doing. It's not about whether they can implement your vision anymore. They can. It's about how well you communicate it.

Good prompts are like good scene direction:

• Clear end goal
• Specific constraints
• Technical requirements
• Expected behavior
• You also have to provide context, this is where the analogy ends 😉

The Productivity Multiplier

When you shift from coding to directing:

• Implement ideas as fast as you can describe them
• Focus on architecture over syntax
• Iterate on concepts, not implementation details
• Ship more, debug faster

The tooling (Cursor, Copilot, etc.) has reached the point where the bottleneck isn't the AI's capability, it's how fast you can articulate what you want.

What I'm building

Delegate tasks. Get software.

Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.

Take a look at vroni.com