liquid-neural-network

LSTN is an experimental text generation engine that models language not as static probabilities, but as a dynamic and "liquid" neural network. Each trigram (3-character sequence) acts as an individual neuron within a continuous temporal dynamic system.

MT-LNN — clean prototype of a brain-inspired liquid neural network (constant-memory recurrent state, O(1) generation cache). Early baseline; active development continues at github.com/everest-an/M1 (AwareLiquid).

Patent-scoped reference implementation for TÜRKPATENT 2026/007632. 17 hazard-based spiking neuron models (TH-LIF, L-TH-LIF, BH-LIF + physics & architecture variants), custom CUDA kernel with analytic natural gradient, and Arty A7-35T FPGA validation (171 LUT, ≤240 pJ/step, bit-exact across 5 references).

SomaOS BrainAgent - a human-brain-inspired embodied AI framework: a full neurocognitive loop over 22 brain regions, with global-workspace arbitration, episodic-semantic memory, endogenous drives, and a perception-cognition-action loop. (Robonix community entry)

Liquid neural networks with input-adaptive time constants: online learning, LFM2 hybrids and a browser demo

Experimental agent pairing a frozen LLM with a plastic liquid neural network that learns online from screen input