How bobertneek/a-life works
A-Life is a research alpha for persistent embodied organisms, with a narrower implemented product boundary: the production voxel frontend starts and ticks GPU-authoritative cognition while the world remains authoritative over legal actions and outcomes. The user-facing loop spans save selection, world perception and candidate generation, GPU cognition, world action execution, and visual projection; the frontend also exposes save/load and inspection controls. The README documents these boundaries, but most application wiring was not sampled, so edges are limited to documented relationships. Semantic and archival capabilities are represented without asserting unverified integration.
Main entry points: organism.rs World state · closed_loop_pipeline.rs Closed-loop dispatch · gpu_live_runtime.rs GPU live runtime · bundle.rs Creature archive
Components 22
Production app
- Production shell
crates/alife_game_app/src/bevy_shell.rs - Voxel frontend
crates/alife_game_app/src/production_voxel_frontend.rs - Voxel renderer
crates/alife_game_app/src/production_voxel_renderer.rs - Camera and inspector
crates/alife_game_app/src/camera_inspector.rs
World simulation
- World state
crates/alife_world/src/organism.rs - World perception
crates/alife_world/src/grounded_sensing.rs - Candidate generation
crates/alife_world/src/candidate_enumerator.rs - Ecology and habitat
crates/alife_world/src/ecology.rs - Bevy sensory adapter
crates/alife_bevy_adapter/src/sensory.rs
GPU cognition
- GPU live runtime
crates/alife_game_app/src/gpu_live_runtime.rs - Closed-loop dispatch
crates/alife_gpu_backend/src/closed_loop_pipeline.rs - GPU brain buffers
crates/alife_gpu_backend/src/buffers.rs - GPU memory
crates/alife_gpu_backend/src/closed_loop_memory.rs - GPU kernel set
crates/alife_gpu_backend/src/closed_loop_v11.rs
Persistence and lineage
- Creature archive
crates/alife_archive/src/bundle.rs - Lineage display
crates/alife_archive/src/legacy_lineage_display.rs - World persistence
crates/alife_world/src/persistence.rs
Learning and semantics
- Training rollout
crates/alife_gpu_backend/src/training_rollout.rs - Lessons and curriculum
crates/alife_school/src/curriculum.rs - Semantic providers
crates/alife_semantic/src/providers.rs - Local SLM prior
crates/alife_semantic/src/local_slm_prior.rs
Other components
- Player
Connections 24
- Player → Production shell: starts game
README.md - Production shell → GPU live runtime: ticks runtime
README.md - GPU live runtime → Closed-loop dispatch: dispatches cognition
- Closed-loop dispatch → GPU brain buffers: uses buffers
- Closed-loop dispatch → GPU kernel set: runs WGSL
- Closed-loop dispatch → GPU memory: updates memory
- GPU live runtime → World state: reads world state
- World state → World perception: grounds perception
- World perception → Candidate generation: generates candidates
- World state → Ecology and habitat: advances ecology
- Bevy sensory adapter → World perception: converts observations
- GPU live runtime → Voxel renderer: projects live state
README.md - World state → Voxel renderer: supplies transforms
README.md - Voxel renderer → Voxel frontend: renders scene
- Player → Voxel frontend: inspects scene
- Voxel frontend → Camera and inspector: opens inspector
- Player → World persistence: saves or loads
README.md - World persistence → Production shell: provides selected save
README.md - Creature archive → World state: imports founders
- World state → Creature archive: archives creatures
- Creature archive → Lineage display: supports lineage display
- Lessons and curriculum → World state: teaches through perception
- Semantic providers → Local SLM prior: offers local provider
- Training rollout → Closed-loop dispatch: trains against runtime
crates/alife_gpu_backend/src/closed_loop_pipeline.rs