A few implementation notes people might be curious about:
• The environments simulate evolving ecosystems rather than fixed tasks.
• Genomes currently use integer-encoded traits with configurable mutation operators.
• I ran large GPU experiments while studying how encoding (binary vs Gray) affects evolutionary stability under mutation pressure.
• The system eventually became a sandbox for long-horizon adaptation experiments.
I’m particularly interested in feedback on mutation models and how people might plug in learning agents.