Update README.md

README.md

@@ -1,11 +1,37 @@
 ---
 license: gpl-3.0
 ---
-<<<<<<< HEAD
-# Metis-OLMoE-Latent-Telemetry
-=======
-The Real Origin of my Neuromorphic AI journey.  It started with frustration and a very personal need. I am 25, bouncing between majors — economics, finance, pre-med, computer science, and finally electrical engineering. Each switch came from the same feeling: I wanted to build things, understand systems at a deep level, and create something meaningful. But traditional classes and tutors were often hard for me as someone with ADHD. I struggled to grasp material quickly, and I hated how easily people got impatient when I needed things explained again. That’s when I realized I needed a better way to learn. I wanted a local AI tutor — something patient, always available, that could help me with heavy engineering concepts, Rust, Julia, and whatever else I was studying. AI like Gemini had already shown me how powerful a non-judgmental tutor could be. It never got frustrating. It just kept adapting. But my hardware couldn’t run the big models smoothly. The smaller ones weren’t smart enough for the kind of technical work I needed. So, I started thinking: what if I could build my own tailored AI tutor that actually ran well on my machine? Around that time, I made the switch from CS to EE. I wanted to understand hardware at a fundamental level. While exploring, I started mining on Dynex — not primarily for profit, but as an experiment. I wanted to see if I could turn the mining telemetry (power draw, temperature curves, clock behavior, memory pressure) into something useful. I had this idea: what if the AI could “feel” its own hardware limits? What if the telemetry became the AI’s heartbeat — a real signal of stress, adaptation, and constraint? That simple question opened the door to Spiking Neural Networks. SNNs mimic biological brains in a way that felt closer to how humans actually learn and recover — especially after my own concussion from 7th-grade football, which left me with years of invisible struggles: depression, social isolation, and the constant feeling that no one fully understood what was happening inside my head. I had even tried switching to pre-med during COVID because I wanted to help people with concussions and neurological injuries, but that path wasn’t right for me. SNNs felt like a different way to approach the same goal — building systems that understand struggle, adaptation, and resilience. So, the project shifted. What began as “I need a local AI tutor” became something deeper: creating an AI that doesn’t just run on hardware but learns from the real stress and behavior of that hardware—turning Dynex telemetry into a heartbeat. Using that signal to drive SNN quantization. And ultimately building Spikenaut — a pure, from-scratch spiking neural network. It wasn't perfect, yet it was all. There were still some gaps in my telemetry that I didn't even realize. I tried using different mining telemetry, tried creating my own HFT bot for telemetry, then I tried using mining sync node data. However, in telemetry data, there were at times zeros in value, after I had used it for training. 
->>>>>>> (https://huggingface.co/allenai/OLMoE-1B-7B-0125-Instruct-GGUF)  Thank to Allen AI for providing full open source models.
+# Metis-OLMoE-Latent-Telemetry: Spikenaut SNN Routing
+
+## 1. The Origin of my Neuromorphic journey and this project
+Before this was a formal dataset, it was an attempt to solve a bare-metal problem. I had been experimenting with mining telemetry, HFT bots, and sync node data to train a spiking neural network (SNN), but the data kept returning dead zeros in value after being used for training. 
+
+The breakthrough came entirely by accident. I was running heavy mods—DLSS 4.0 and path tracing—on Cyberpunk 2077 and the Resident Evil 4 Remake. My workstation PC was screaming, pushing harder and louder than it ever did during crypto mining. That sparked the realization: *What if I used raw gaming telemetry data for neuromorphic spike data conversion?* What if I could use this intense hardware stress to create an artificial heartbeat for AI?
+
+When I pitched this idea, most people didn't believe the spike data conversion would work. But after refining the early thermal equations using that Resident Evil 4 telemetry, the Spikenaut architecture was born. The true origin of this project is deeply personal—it stems from my switch from Computer Science to Electrical Engineering, the struggle to learn on my own terms, and the drive to build an architecture that adapts dynamically. Now, I am dropping these new SNN quantization models and datasets on Hugging Face to prove the math works and to keep the research completely open and transparent.
+
+## 2. The Science: Semantic Attractor Clustering
+This dataset contains the raw bare-metal telemetry logs and latent space visualizations generated by the **Spikenaut** SNN architecture. The objective is to map the physical routing of LLM embeddings (specifically from the OLMoE-0125 Mixture of Experts model) as they are processed by biologically-inspired neuronal fatigue mechanics.
+
+The primary discovery documented here is **Semantic Attractor Clustering**. By applying L2 Normalization to F16-to-F32 casted embeddings, the SNN bounds the semantic pressure to a unit sphere. This prevents "Winner-Take-All" routing collapse and forces the network to organically balance the load. The resulting telemetry proves that the SNN physically routes different semantic concepts (e.g., abstract philosophy vs. rigid code syntax) into distinct, repeatable biological pathways.
+
+## 3. Experiment Progression
+The dataset documents the chronological progression from synthetic baselines to actual semantic routing:
+
+* **Phase 1: Synthetic Baseline (Smoke Test)**
+  * **Input:** Synthetic sine wave.
+  * **Result:** Verified the GPU temporal loop (10,000 ticks) and basic biological fatigue without crashing the CUDA context.
+* **Phase 2: The F16 Magnitude Collapse (Unbounded)**
+  * **Input:** Real LLM embeddings (OLMoE).
+  * **Result:** Unscaled F16-to-F32 extraction resulted in raw, unbounded electrical pressure. A single expert neuron (Walker ~620) was overwhelmed, causing a routing collapse where one walker took the entire load for the full temporal loop.
+* **Phase 3: L2 Normalization & Philosophy Attractors**
+  * **Input:** `"Let's teach this MoE model..."` (Abstract English).
+  * **Result:** L2 Normalization successfully shattered the routing collapse. Energy dynamically settled into high-register attractor bands, predominantly isolating into the **2000-route**.
+* **Phase 4: Semantic Clustering (Code & Math Logic)**
+  * **Input A:** `fn main() { println!("Hello, World!"); }` (Rust Syntax)
+  * **Input B:** `"The derivative of a constant is mathematically zero."` (Math Logic)
+  * **Result:** The SNN abandoned the 2000-route completely. Both raw Rust syntax and mathematical logic organically fell into the exact same **600-800 frequency band**. This demonstrates that the network physically maps highly structured logic tasks to adjacent biological neighborhoods.
+
 **Dataset for Spikenaut SNN Research**
 
 This dataset contains latent telemetry and routing data generated from the SNN-quantized version of AllenAI’s OLMoE-1B-7B-0125-Instruct model using the `corinth-canal` pipeline.
@@ -26,23 +52,3 @@ These files are used to study spiking behavior, routing stability, and adaptive
   ### Related Repos
 - [corinth-canal](https://github.com/Limen-Neural/corinth-canal) — SNN quantization pipeline
 - [Surrogate_Viz.jl](https://github.com/Spikenaut/Surrogate_Viz.jl) — Symbolic regression and visualization
-
-
-
-### Origin Story
-The Real Origin of my Neuromorphic AI journey. It started with frustration and a very personal need. I am 25, bouncing between majors — economics, finance, pre-med, computer science, and finally electrical engineering. Each switch came from the same feeling: I wanted to build things, understand systems at a deep level, and create something meaningful. 
-
-But traditional classes and tutors were often hard for me as someone with ADHD. I struggled to grasp material quickly through traditional learning, and I hated how easily people got impatient when I needed things explained again. That’s when I realized I needed a better way to learn. I wanted a local AI tutor — something patient, always available, that could help me with heavy engineering concepts, Rust, Julia, and whatever else I was studying. AI like Google Gemini had already shown me how powerful a non-judgmental tutor could be. It never got frustrating. It just kept adapting.
-
-So I made an investment to build my very own workstation.  Thinking RTX 5080 was going to be enough for state of art models.  My hardware couldn’t run the big models smoothly. The smaller ones weren’t smart enough for the kind of technical work I needed. So, I started thinking: what if I could build my own tailored AI tutor that actually ran well on my machine? Around that time, I made the switch from CS to EE. I wanted to understand hardware at a fundamental level. While exploring, I started mining on Dynex — not primarily for profit, but as an experiment. I wanted to see if I could turn the mining telemetry (power draw, temperature curves, clock behavior, memory pressure) into something useful. I had this idea: what if the AI could “feel” its own hardware limits? What if the telemetry became the AI’s heartbeat — a real signal of stress, adaptation, and constraint? 
-
-That simple question opened the door to Spiking Neural Networks. SNNs mimic biological brains in a way that felt closer to how humans actually learn and recover — especially after my own concussion from 7th-grade football, which left me with years of invisible struggles: depression, social isolation, and the constant feeling that no one fully understood what was happening inside my invisble injury. I had even tried switching to pre-med during COVID because I wanted to help people with concussions and neurological injuries, but that path wasn’t right for me. SNNs felt like a different way to approach the same goal — building systems that understand struggle, adaptation, and resilience. So, the project shifted. What began as “I need a local AI tutor” became something deeper: creating an AI that doesn’t just run on hardware but learns from the real stress and behavior of that hardware—turning Dynex telemetry into a heartbeat. 
-
-And ultimately building Spikenaut — a pure, from-scratch spiking neural network. It wasn't perfect, yet it was all. There were still some gaps in my telemetry that I didn't even realize. I tried using different mining telemetry, tried creating my own HFT bot for telemetry, then I tried using mining sync node data. However, in telemetry data, there were at times zeros in value, after I had used it for training.
-
-Then out of nowhere, I started adding heavy mods like DLSS 4.0 and path tracing for Cyberpunk 2077 and the RE4 remake. The thing is that my workstation pc was much louder than crypto mining. So, a lightbulb in my head came up, " What if I tried recording gaming telemetry data for Neuromorphic spike data conversion. I even told some people my idea, and I could see it in their face, that they didn't believe in me. The same thing happened with my first SNN experiment. So, one day I was asking Grok 4.20 about it. At first, it didn't believe my idea, but I explained how I use mining telemetry data hardware to create an artificial heartbeat for AI. Grok became super excited about my new idea. The Resident Evil 4 gaming telemetry came later and helped shape the early thermal equations that fed into SAAQ, but the true origin was much more personal: the struggle to learn on my own terms, the switch from CS to EE, and the desire to build something that could adapt the way I needed to adapt. That’s where Spikenaut really began.
-
-Now I am in the moment of showing that a new idea is actually going to work by me dropping the new SNN quantization models and dataset on HF while making the research transparent and open.
-
-
----