What is nano-neuron?

NanoNeuron is a teaching project that shows how machine learning actually works under the hood, using just seven short JavaScript functions and no external libraries. The goal is to strip away the mystery and show that a neural network is, at its core, math, not magic. The example is intentionally simple: a single artificial neuron learns to convert temperatures from Celsius to Fahrenheit. It starts with random guesses, then gradually improves by measuring how wrong it is and adjusting two internal numbers (called weight and bias) after each round. The README walks through every step in plain language, including how errors are calculated, how the program figures out which direction to adjust those numbers, and why repeating the process thousands of times leads to accurate predictions. The four main ideas the code demonstrates are: making a prediction, calculating how far off that prediction is, figuring out how to nudge the internal numbers in the right direction, and repeating all of that in a training loop. After 70,000 training rounds, the neuron ends up with values very close to the actual formula for the conversion, even though it was never told what those values should be. The README is honest about what is left out. A single neuron is much simpler than a real neural network. Several real-world techniques, like normalizing inputs, using matrix math for speed, or applying activation functions, are skipped to keep things readable. The project is meant as a starting point for someone who wants to understand what is happening inside these systems before moving on to larger frameworks. You can run the code locally by cloning the repository and running a single Node.js command. No installation of additional packages is required.