ClearDataLabs
Neural Networks Explained — Interactive Demos in the Browser
Complex AI concepts, explained clearly and simply. Interactive experiments, in-depth articles, and hands-on projects — explore how AI actually works, from the math up.
Recent Articles
-
The Architectural Loophole: AI Copyright & Book Replication
Discover how AI refuses copyrighted PDFs but recreates entire book structures from memory. Explore the ethics of Architectural Replication and synthesized displacement.
-
What ClearDataLabs Is About
ClearDataLabs is an open-source project that explains AI and neural networks through interactive browser demos and in-depth articles — building everything from scratch, no ML libraries.
-
Seeing in Cells: How Handwriting Becomes Input for a Neural Network
hwrjs
How to preprocess handwriting for a neural network: raw pen strokes become a scale-invariant 12×12 binary grid of 144 numbers — implemented in TypeScript, no libraries.
-
Making a Neural Network Dream: DeepDream-Style Visualization in JavaScript
knowledge-extract
A DeepDream-style technique in pure JavaScript: optimize a blank image until a neural network confidently sees a digit. Visualize what an MNIST network has learned.
-
What Does a Neural Network Learn? Visualizing MNIST with Causal Index
knowledge-extract
What has a neural network trained on MNIST actually learned? We compute a causal index per pixel and visualize the result as a heat map — implemented in pure JavaScript.
Featured Projects
Handwriting Recognition in the Browser
A neural network that reads your handwriting — trained and running entirely in JavaScript, no server, no ML library. Draw a letter and watch it classify in real time.
Knowledge Extraction from a Neural Network
What does a network trained on MNIST actually learn? Two methods to find out: causal index heat maps and iterative input optimization — both running in pure JavaScript.