ALLA doesn't use datasets, pretrained models, embeddings, or LLMs. It learns by asking, understanding, building concepts, and reflecting — just like a child. The system is entirely symbolic, recursive, and modular.
What ALLA does: Learns new concepts through user interaction
Builds its own semantic network (concepts and relations)
Has self-generated goals (e.g., UNDERSTAND, REMEMBER)
Reflects on its memory to form abstractions and analogies
Grows knowledge recursively — from 0 → 1M semantic connections in ~1 hour
ALLA's architecture includes 7 modules: Command Parser, Semantic Memory, Planner, Execution Engine, Goal Manager, Reflection Cycle, and Semantic Bootstrapper.
No Copilot suggestions dictated structure. I guided every logic block, using Copilot only for syntax. This project is 100% solo — from engineering to philosophy.