Observe.
Capture the rhythm of a fight. Movement, timing, spacing, and the decisions that shape each exchange.
CONTEXT IN MOTIONA learning system built for the arena.
EVO studies combat, tests its decisions, and turns experience into the next iteration.
PROJECTEVO AIBY ALATIC
RESEARCH DOMAINCOMBATMINECRAFT JAVA
POLICYLEARNEDTRAINED WEIGHTS
PUBLIC RESULTSAWAITING DATANO PUBLISHED BATCH
MODEL RELEASESREVIEWEDMANUAL PROMOTION
One continuous research loop.
Progress is measured, not assumed.
Capture the rhythm of a fight. Movement, timing, spacing, and the decisions that shape each exchange.
CONTEXT IN MOTIONTurn recorded experience into candidate policies. Compare each iteration against the model that came before it.
EXPERIENCE INTO WEIGHTSTest what changed. Keep the evidence. A new candidate earns its place through evaluation and review.
REVIEW BEFORE RELEASEA new model is a hypothesis. Recorded experience informs the next candidate; evaluation determines what actually improved.
OUR PRINCIPLES ↗Verified matches only
Aggregate combat results
Published evaluation sample
Only selected aggregate results are published. No player identities, recordings, model files, or internal configuration.
No verified public batch has been published yet. Empty metrics are intentional. The core animation is illustrative, not a live performance feed.
The goal is better decisions.
The standard is real evidence.
Combat policy comes from trained weights. Progress means improving decisions, not dressing up a fixed routine.
A small public window into the project. Research data and control tools stay inside the owner-only control room.
No invented ranks. No unbeatable claims. Candidate models are evaluated before any manual promotion.