Hey, I'm Raul. I'm an out‑of‑the‑box tinkerer, builder, and relentless experimenter based in San Marcos, Texas.
What hooked me was the elegance — the way spikes, timing, and sparse events could mirror real brain‑like computation. The excitement was addictive, and suddenly I was in deep hyperfocus: experimenting with temporal coding, data extraction, playing with FPGA ideas, and trying to build biologically plausible systems from the ground up, right on my new build rig.
Rapid growth + hyperfocus = messy early codebase. I used GitHub like a cloud backup, not a development platform. Accidental deletions wiped chunks of work. Modules scattered. Dependencies baked into my local Fedora setup. Everything is now moving toward being portable, reproducible, and free of local environment quirks.
- Removing Fedora‑specific and other local dependencies
- Eventually bringing back HDL libraries for FPGA neuromorphic enthusiasts
- Spikenaut will return in a cleaner, stronger form as I rebuild the algorithms, data engineering, and architecture behind it.
This won't happen overnight — it's a deliberate, careful rebuild. But I'm committed to doing it right.
I want Limen Neural to become a clean, open, community‑friendly project — not something trapped on my local machine. I want to rebuild the parts that were lost, refine the parts that survived, and finally share the work I’ve been doing on Spikenaut, neuromorphic data and algorithms, now SAAQ (Spiking Activity Adaptive Quantization) and Metis (MoE‑SNN), in a way that others can actually use.
Even as an opportunity to learn from y'all.
If you have ideas, suggestions, or want to collaborate, I’d genuinely love to hear them.
The long‑term goal is simple:
Limen Neural provides a clean, reusable foundation for:
-
Spiking Neural Network (SNN) encoding
Rate, temporal, population, and neuromodulated encoders. -
GPU‑accelerated SNN simulation
Designed to integrate with CUDA, maybe even ROCm, or CPU backends. -
Telemetry extraction & quantization
Including SAAQ (Spiking Activity & Adaptive Quantization) and data extraction for hardware‑driven learning. -
LLM ↔ SNN fusion research
A standardized interface for converting embeddings, latents, or activations into spike‑based dynamics.
This future library is going to be the “generalized chassis” extracted from the original workstation‑bound architecture — now being rebuilt to be portable, reproducible, and open.
This project is as much about the journey as the destination. Thanks for stopping by — let's push neuromorphic computing forward together, one spike at a time. ⚡
“The long‑term goal is simple: Build a modular, hardware‑agnostic toolkit for encoding, simulation, telemetry, neuromorphic algorithms, SNN‑LLM quantization, and bio‑inspired computation — without relying on any local environment quirks.”