I’m working on an AGI-oriented architecture project called Geniea and looking for technical review + possible collaboration.
The idea is to explore a more grounded path beyond pure LLM-agent systems by combining perception, symbolic knowledge, world models, concept formation, neuro-symbolic reasoning, continual learning, and hard safety constraints.
I’m not claiming AGI is solved — I’m trying to pressure-test whether this architecture has real potential and where it breaks.
Would especially love feedback on:
novelty,
feasibility,
technical bottlenecks,
benchmark design,
and relevant prior work.
Also open to collaboration with people interested in world models, neuro-symbolic systems, alignment, verification, and agent design.
Strong criticism is very welcome — I’d rather find the weak points early than sit inside my own bubble.
Happy to share the architecture doc if anyone wants to review it critically.