openfluke01 / EXPLORE & RECORD
Make a trail.
Walk or fly a route and record it. That journey becomes a demonstration for a Spark, one of the little creatures in your world.
ALPHA · AN AI EXPLORATION SANDBOX
A world to explore.
An AI to teach.
What if your little companion could learn from what you do—and remember it next time?
ALife: Becoming is a game about exploring, experimenting, and helping artificial life learn through experience. Built around continuous on-device learning and persistent memory, it brings the AI research into a world you can play in.
EXPLORE TOGETHER. LEARN THROUGH PLAY.Playable alphaStill growing toward betaMultiplayerCross-platform web & local playLearning that staysSaved AI models and trails
A LOOK INSIDE THE ALPHA
Can my game AI
learn and remember?
A real development recording of TwinTrace in ALife: Becoming.
The experiment: learn from a stream of gameplay, then try the learned behaviour after the teacher is removed.
Watch on YouTube ↗THE IDEA, WITHOUT THE JARGON
You show it. It learns.
What can it remember?
02 / TEACH
Learn from the journey.
Train the Spark on the trail. TwinTrace pairs a model that adapts to new experience with another that holds on to what has been learned.
03 / REMEMBER
Let it try for itself.
PLAY AI uses the saved model to guide the Spark, with the recorded teacher switched off. Reopen the game and return to a trained trail to try its memory again.
Learning on your own hardware.
The game is built around AI that can keep learning from experience on the device running the simulation. In local play, that is your machine; in a hosted multiplayer world, the host runs the simulation and learning.
A remembered skill, not just a replay.
Playing a recorded trail shows the original movement. Playing AI asks the learned model to act. Comparing the two is part of the experiment: how much useful behaviour survives without the teacher?
WHERE THE GAME IS NOW
In alpha.
Becoming something more.
The core game loop is still being finished. Beta is the next milestone, not the current release.
A world you can already explore.
- Cross-platform multiplayer through web and local play.
- Connected planetary worlds, trio forms, shared physics, skills, and the beginnings of crafting.
- Record trails, train Sparks with TwinTrace, and run the saved AI model.
- Persistent trails and learned models, with trained Sparks restored when you return nearby.
Give that learning a purpose.
The intended loop is to do a job yourself, teach it to a Spark, and put that learned behaviour to work—then use what you gather to build, travel, and experiment further.
Finishing that loop, progression, and the wider sandbox is ongoing work. The current learning demonstration is a step toward it, rather than a finished automation game.
FROM PROTOTYPE TO GAME
ALife OS was the prototype.
Becoming is the game.
The experiments in ALife OS laid the groundwork. ALife: Becoming carries that work into a dedicated game, with its own multiplayer foundation and the new TwinTrace learning experiments.
Welvet is the AI engine underneath. OpenFluke is the independent work of Samuel Watson bringing the engine, research, and game together.