← All Demos
robert@barcik.training

Neuron Playground

A neuron is the fundamental building block of neural networks. It takes multiple inputs, multiplies each by a learnable weight, adds a bias, then passes the result through an activation function to produce an output. Adjust the sliders below to see how each parameter shapes the neuron's behavior in real time.

Inputs & Weights
Input x1
I see steam rising
Steam doesn't always mean danger — we see steam from hot showers and kettles. Slightly positive weight.
x1 0.0
w1 0.4
0.0 × 0.4 = 0.00
Input x2
The burner is glowing red
A glowing red burner is strong visual evidence of extreme heat. High positive weight — this is a reliable danger signal.
x2 0.0
w2 1.5
0.0 × 1.5 = 0.00
Input x3
Someone says it's cool
Someone saying "it's cool" should reduce the danger signal — but people can be wrong. Negative weight, but not too strong.
x3 0.0
w3 -0.6
0.0 × -0.6 = 0.00
Bias
b 0.0
0.4 1.5 -0.6 x1 0.0 x2 0.0 x3 0.0 σ output 0.00
Output
Weighted Sum
0.4·0.0 + 1.5·0.0 + (-0.6)·0.0 + 0.0
= 0.00
Activation Function
ReLU(x) = max(0, x)
Neuron Output
0.00
0 Signal Strength max
Analogy: Weights as Volume Knobs

Think of weights as volume knobs — they control how much the neuron “cares” about each input. A large positive weight amplifies the signal; a negative weight inverts it; zero means the neuron ignores that input entirely.

Scale: From One Neuron to Trillions of Weights

In a real neural network: today's frontier models are estimated to have on the order of a trillion or more weights (exact figures are no longer disclosed — GPT-4 was famously rumored at ~1.8T back in 2023). Each weight was learned during training by processing trillions of text tokens. This single neuron demonstrates the fundamental building block that, when combined in layers of millions, gives rise to language understanding.

AI transparency

This demo is a scripted, self-contained browser simulation. Nothing you type is sent to an AI model and no live AI system runs behind it, even where it plays one. Its code and copy were built with generative AI (Anthropic’s Claude) and reviewed by Robert Barcik, who is responsible for what is published (LearningDoe s.r.o.). Disclosed in the spirit of Article 50 of the EU AI Act.