Bots that learn
Opponents are trained in a Python reinforcement-learning stack and run in-engine as ONNX policies. They improve by playing, not by hand-written rules.
Artificial Environment Lab is a cross-platform multiplayer shooter built in a custom Godot 4 engine. Its bots are trained with reinforcement learning, not hand-scripted, and every build is a real, signed release you can download and play now.

A lab first, a shooter second: one arena for human players and the machines learning to beat them.
A research facility built for combat: reactor halls, mezzanines and connecting spans, with cover and sightlines authored for fast matches. It's the first finished map; more arenas are in development.

Every part of AE//LAB is built in-house: the engine fork, the networking layer, the bot runtime and the training stack.
Opponents are trained in a Python reinforcement-learning stack and run in-engine as ONNX policies. They improve by playing, not by hand-written rules.
A custom Aether networking layer handles live replication, reliability and matchmaking, not Godot's default multiplayer.
Matches record to a compact .aerx format and play back exactly: the same input produces the same result.
Every match resolves to a scored result that feeds a local leaderboard. The outcome is computed from the match.
Keyboard and mouse, gamepad, or on-screen touch controls with a virtual joystick. One build handles every input.
A bolt rifle, marksman rifle, pistol, mortar, lance, blade, spike and bouncer, each with its own handling and range.
The same signed release ships to desktop and mobile, and everyone plays in one matchmaking pool. Free, and no account required to download.
64-bit desktop, delivered as a portable .zip. Unzip and run.
64-bit desktop build as a portable .zip for modern distributions.
A direct-install .apk (arm64), or install from Google Play.
Free, direct download for Windows, Linux and Android. Real signed builds, checksummed, no launcher and no account.