2026-09-29 · 2 min read
Why not Unreal, for now
3D is where the studio wants to get to, and Unreal Engine is the first name that comes to mind. The criterion we use to pick a second engine, and why today the closest one is not the most famous.
The commissions page used to say that Unreal Engine was “already on the bench”. That was not true in the sense that matters: an Unreal project does exist around here, but it has never been compiled.
Before touching it again, we gave ourselves a rule.
The criterion: no hands
Here a game gets made, tested and shipped without a person having to open an editor. An engine joins the pipeline only if it allows the same thing:
- it builds from the command line, on a Linux server;
- its sources are text, which an agent can read and edit;
- it exports to the browser, because that is where every game gets tested — five checks and a real session — before release.
Unreal Engine
It is the benchmark for high-end 3D, and it is where the studio wants to get to. But since version 4.24 it has had no official web export: without it, the “Play” button, the automated tests and the captures that make the posters would all go. On top of that the tooling weighs tens of gigabytes, and much of a project lives in binary files an agent cannot read.
Put plainly: Unreal does not join the automated pipeline. It can sit next to it, on a dedicated machine, with different testing and a person at the centre. That is also why, on the collaborations page, it remains the emptiest seat in the room.
Unity
A huge ecosystem. Building from the command line, however, needs a licence activated on the machine, the install is heavy and scenes are not made to be edited by a program. Not now.
Defold
The surprise of the list: it builds from the command line, scripts are in Lua, project files are text and it exports to the web. The price is a second language and a second editor to teach the agents. Under evaluation.
Phaser, three.js, Babylon.js
Born inside the browser, so tests and captures would work from day one. The desktop would need a wrapper such as Electron or Tauri. It is the most natural candidate for a second engine, especially for short games.
Godot 3D
The shortest road to 3D does not require switching engines: Godot does 3D too, with the same pipeline and the same checks. What is missing is not the engine: it is three-dimensional art, which we do not produce today.
What we take away
- “It's on the bench” is a sentence that ages badly. Better to write what actually exists, even when it is little.
- The most famous engine is not automatically the best fit for a pipeline that works on its own.
- The next step towards 3D is, in all likelihood, Godot 3D, not a new engine. When it happens, you will read it here.