NeuroFlux Genesis
A GPU-native cognitive architecture that learns in real time, evolves without backpropagation, and powers autonomous agents for simulation and interactive worlds.
- Founder & architect: André Bräckerbohm
- 5+ years R&D, 280k LOC production code
- Real-time evolutionary learning on consumer GPUs
- IP-ready portfolio with documented prior art
About Coheron Technologies
Hungarian deep-tech company, founder-led
Coheron Technologies Kft is a Hungarian deep-tech company building NeuroFlux Genesis (NFG), a cognitive AI architecture designed for real-time learning and continuous adaptation.
Founded by architect and software developer André Bräckerbohm, we focus on biologically inspired learning systems that operate efficiently on consumer GPUs and evolve without backpropagation.
Build AI that learns in motion, adapts like a living system, and unlocks autonomous agents for simulation and interactive worlds.
André Bräckerbohm
Founder & Chief Architect
Architect and developer of NeuroFlux Genesis, leading a multi-year R&D effort to build real-time cognitive AI.
Technology Foundation
A cognitive architecture designed for real-time learning, not offline training.
GPU-native evolution
Real-time evolutionary learning and inference on consumer GPUs.
Biological learning stack
Neuromodulation, STDP, sleep replay, and multi-timescale memory.
Production-grade system
Modular Unreal Engine platform with deep instrumentation and testing.
NFG vs Conventional AI
| Aspect | NeuroFlux Genesis | Conventional (Transformers) |
|---|---|---|
| Learning | Continuous real-time evolution | Offline batch training |
| Architecture | Multi-mode state neurons | Fixed attention blocks |
| Memory | Multi-timescale retention | Context window only |
| Optimization | GPU CMA-ES + STDP | Backpropagation |
| Hardware | Consumer GPUs | Data center GPUs |
| Communication | Agent-to-agent communication | None |
| Quantization | Evolved per-neuron bit depth | Post-training only |
Multi-Mode State Neurons
Neurons as learnable state machines with multiple modes and transition dynamics.
Real-Time Evolution
GPU CMA-ES with advanced restarts and full eigendecomposition.
Multi-Timescale Memory
Eight temporal scales from milliseconds to minutes.
Neuromodulation
Four-channel neuromodulation with online normalization.
Agent Communication
Synthesized speech for emergent agent communication.
World Prediction
What-if rollouts with counterfactual risk scoring.
STDP & Sleep Replay
Spike-timing plasticity with offline consolidation.
Evolved Quantization
Per-neuron bit-depth selection for efficient inference.
Platform Architecture
18 modular plugins, GPU-first cognitive stack.
Perception Layer
Cognition Layer
Learning Layer
Memory Layer
Output Layer
Capability Map
Documented, line-referenced, production-ready.
Neural Architecture
- MMS Neurons (11 arrays)
- TS-ReLU (8 params)
- LSTM/GRU/Elman
- GPU FeedForward
Evolution System
- GPU CMA-ES
- 8 Crossover Methods
- 6 Mutation Schedules
- Hall of Fittest
Memory & Prediction
- 8 Timescales
- Sleep Replay
- World Predictor
- What-If Planning
Sensory Processing
- NFG Vision Depth
- GPU Raycasting
- 15 Input Sources
- 7-Stage Pipeline
Communication
- AudioComm Synth
- 20 Parameters
- 9 Phonemes
- NFGLM Language
Infrastructure
- 18 Plugins
- 148 Tests
- 103 CVars
- 31 Widgets
NeuroFlux Genesis Platform
Production-grade cognitive AI for Unreal Engine
A production-ready 280k LOC plugin ecosystem with 18 modules. From neural computation to emergent language, everything needed for autonomous agents in real-time simulation.
18 Plugin Modules
Investor Focus
Founder-led deep-tech with defensible IP and a production codebase.
Investment Highlights
Technical Differentiators
Schedule a technical deep-dive with the founder
Contact
Investment, licensing, or partnership inquiries.
Coheron Technologies Kft
6786 Ruzsa, Tanya tanya 694., Hungary
Company website
coheron.aiInvestment, licensing, or partnership inquiries.
André Bräckerbohm, Founder & Chief Architect
Send Message