What is ssnn?

ssnn is a small experiment in building spiking neural networks, a type of brain-inspired AI that's quite different from the neural networks most developers use today. Instead of treating neurons as number-crunching units that output continuous values, spiking neurons communicate through discrete electrical pulses (spikes), much closer to how biological brains actually work. The project implements two key ideas from neuroscience researcher Izhikevich. First, a mathematical model of how individual neurons decide when to fire. Second, a learning rule called STDP (Spike-Timing-Dependent Plasticity) that adjusts the connection strength between neurons based on the timing of their spikes, if neuron A consistently fires just before neuron B, their connection strengthens. The README describes a twist on the standard STDP formula where weights update directly rather than through a derivative. The author walks through how to use this to recognize handwritten digits. You'd convert a dataset like MNIST into JSON, show images to the network for set time intervals, let STDP learning adjust the connections automatically, then freeze the weights and test how well specific "label" neurons respond to each digit. The workflow is essentially: feed in signals, let the network learn on its own, then lock the learning and see how it performs. Who would use this? Probably someone curious about alternative approaches to neural computation, or a student learning about neuromorphic computing concepts. It's written in JavaScript, which is unusual for this kind of work, most spiking network research happens in Python or specialized frameworks. The project is clearly a learning exercise rather than a production tool, with open questions flagged throughout and only a couple of stars on GitHub. It's an accessible starting point if you want to understand spiking networks by reading code rather than papers.