15:I[981,["658","static/chunks/658-a698693f56d0116a.js","183","static/chunks/183-6beeeb801c274aaa.js","474","static/chunks/app/%5Btype%5D/layout-631c4730b83c8223.js"],"Posts"] 16:T47a,Hello HN,

I built Syna to understand how modern ML frameworks like PyTorch actually work — from the ground up.

It’s a minimal, define-by-run (dynamic graph) framework inspired by DeZero, written entirely with NumPy. Unlike most libraries, Syna includes a basic reinforcement learning module right inside the same framework — no separate packages.

It’s not about speed or GPUs — it’s about clarity, simplicity, and learning the internals of machine learning. Great for students, educators, and anyone curious about what’s really happening under the hood.

GitHub: https://github.com/sql-hkr/syna

I also built a web app that visualizes how neural networks learn in real time — perfect for beginners exploring training dynamics:

GitHub: https://github.com/sql-hkr/xor Demo: https://sql-hkr.github.io/xor/

Happy hacking!17:T4aa,I built RIHU (Retrieval In the Hypothetical Universe) while wondering: how could we represent the world we live in as data? RIHU is a small open-source prototype that explores what retrieval might look like beyond RAG — when knowledge becomes geometry, not just vectors. Instead of embedding similarity search, RIHU constructs a geometric “universe” of Hypothetical Facts, Classes, and Instances. Each concept exists along interpretable spatial axes (like time, location, or semantics), and retrieval becomes wayfinding in that space.

Key ideas: - Geometry over similarity: proximity = relevance, density = salience, containment = context - Two retrieval modes: - Objective (global): find centroids a

Tracking Down a Regression in Mesa 3D