This post was originally written in Italian by Paolo Gambardella. It has been translated by an AI agent and may contain inaccuracies.
Prototypes are one of the areas where game design has evolved enormously over the past few months. New machine learning and automation tools mean everyone can move much faster when building the artefacts we need to learn from — and to help our team learn. That’s a real step forward, but it can amplify a tendency I’ve noticed in some teams: building a prototype that’s too polished and ends up becoming a demo.
A prototype exists to help you learn, to answer questions. A demo exists to convince people, to get a greenlight, to sell. The problem appears when you create something refined enough to look convincing, but too ambiguous to tell you whether you actually learned anything. You end up neither learning nor convincing anyone.
Let me give a concrete example. If this is your first time reading this, you should know I’m a game design consultant specialising in concepting and pre-production for new video games, primarily free-to-play. I run a small studio, and last week my team and I built this prototype using Godot:
- The project kicked off on Monday with a vision I’d been carrying around for a while, which I then shaped into a vision document
- My assistant Eduard researched other games and handed me a concept document on Wednesday
- On Thursday I fed that document to Claude Code and asked it to generate a basic prototype
- From Friday to Sunday I made small tweaks based on ideas that came to me while I was doing other things.
After handing the result to my assistant, he played it for a whole day and gave me his take. The conversation drifted heavily towards things like “the generated track is monotonous” or “you can’t really collide with the other cars” — all valid concerns from demo stage onwards, but not at this point. The two of us, talking it through, managed to identify the real learning: what we actually like about this prototype. We set it aside, document it, and move on to the next learning phase.
If it had just been me, or if we’d been at a bigger company, we’d probably have also needed to put together a presentation of the learnings to get everyone aligned. That’s how I like to work — but I’ll be honest: the temptation to fix the technical issues or improve the AI of the enemy vehicles was real. Why? Because right now it’s easy. It’s quick.
You can do it, but so can your competitors. The way I see it, the winners are the ones who focus on the right things at the right time. Discernment matters more than ever today. A contact of mine mentioned that App Store review times are getting significantly longer, given the volume of vibe-coded apps now landing in moderators’ queues. If we’re going to wait longer for approval anyway, we might as well be waiting on something genuinely worth shipping.