Era

I built Era because I was frustrated with LLM hallucinations and the high cost of inference.

Era is an experimental cognitive architecture that replaces neural networks with deterministic math: 1. "Truth" is calculated using Force-Directed Graphs (consensus attracts, outliers repel). 2. "Mood" is simulated using Lorenz Attractors (chaos theory). 3. "Memory" is mapped onto a 3D Toroidal topology.

It runs locally on CPU, costs $0/query (excluding search API), and in my benchmark of 120 queries, it achieved 99% accuracy on fact verification with zero hallucinations.

It is obviously not a replacement for LLMs in creative writing, but for "grounded" tasks (verification, causal reasoning), it offers a transparent, lightweight alternative.

The core engines (1-8) are open source (Apache 2.0). I'd love your feedback on the architecture.