Creating a video game prototype often starts with a simple idea, but turning it into something playable can require hours of adjustments. Playco says GPT-6 Astra is reducing some of that work by helping its teams build, test, and improve games directly inside engines such as Unity and Godot.
Three games from the same foundation
The Playco team started with a non-themed prototype built from basic shapes and elements. This first version worked as a kind of playable sketch, known in development as a grey box.
From that foundation, GPT-6 Astra helped create three prototypes with different themes. According to Playco, all three were generated in a single run, and most worked on the first attempt.
What’s the practical advantage? Developers can compare more ideas in less time. Instead of imagining how each concept would feel, they can play it and decide which one deserves to move forward.
If you have 10 ideas for a game, you can make all 10, play them, and see how they feel instead of limiting yourself to imagining them.
Fewer manual fixes for developers
Playco says GPT-6 Astra reduced manual fixes by 50% compared with the previous model. With the earlier version, the initial prototype needed so many adjustments that continuing to fix it through instructions became unproductive.
In those cases, engineers had to step in directly to repair the game. With Astra, the company says the first version was much more solid, and most of the changes made were related to design and gameplay preferences rather than fundamental errors.
One of the prototypes, which had a cyberpunk aesthetic, did need a performance fix. The others, according to Playco, were completed without any significant additional rounds of iteration.
AI doesn’t just write code
Playco is using GPT-6 Astra as part of Playbot, an AI development environment designed for video game professionals. The tool connects to engines such as Unity and Godot so the model can edit scenes, run games, test changes, and detect problems within the usual workflow.
This matters because developing a game involves much more than generating lines of code. AI also needs to understand where to place objects, how an interface will respond in different situations, whether a reference image was recreated correctly, and whether a change actually improves the playing experience.
According to Joao Vieira, Playco’s product engineering lead, Astra showed improvements in spatial reasoning, reproducing visual references, and creating more responsive interfaces within Unity. The company also observed progress in what is known as game feel—the sensation a game conveys as it moves, responds, and reacts to the player’s actions.
Test, detect, and improve
Playbot allows the model to play and validate its own changes. This helps it find errors more easily and identify parts of the experience that still need improvement.
This approach changes AI’s role in development. It is no longer limited to producing a piece of code and waiting for someone to check whether it works. It can also take part in a more complete cycle: build, run, observe, fix, and test again.
Even so, human judgment remains essential. Playco explains that some adjustments were made based on gameplay preferences—a decision that depends on more than whether the game works without errors. A prototype can be technically correct and still not be fun.
The real impact is being able to experiment more
The most interesting promise of this progress is not creating complete games without human involvement. It is allowing teams to turn more ideas into playable prototypes before investing large amounts of time in a single option.
For small studios, independent designers, and professional teams, that speed can make creative exploration easier. AI handles much of the repetitive work while people decide which concept has potential, which mechanics work, and what kind of experience they want to build.
GPT-6 Astra does not eliminate the complexity of video game development, but it can shorten the distance between an idea and a first playable version. And when testing an idea costs less, there is also a greater chance of discovering a good idea that might otherwise have remained in a notebook.