Mister Atompunk Presents: Atomic Almanac Season Zero

Level 4 — THE OPTIMISM MACHINE

Perceptron (1957)

"Electronic brain teaches itself!" — Cornell Daily Sun, 1958

What this is

An interactive re-creation of the Perceptron — Frank Rosenblatt’s single-layer neural network, once hailed as the machine that could learn anything. Optimism soared: the age of intelligent machines had arrived. Then came XOR, the proof of limits, and the first AI winter.

Playthrough

Phase 1 — The Machine That Thinks
Train a perceptron on simple shapes. Circles vs squares, perfectly separated by a line.

Phase 2 — Linear Mastery
Place your own training points. Watch the algorithm converge, drawing a decision boundary.

Phase 3 — The XOR Trap
Confront the fatal flaw: a pattern no line can separate. Optimism fades as failure repeats.

Phase 4 — Winter Descends
Funding collapses. Labs shutter. The first AI winter falls, even as whispers of deeper networks remain.

Controls

• Click to place training points

• Buttons to select class A or B

• Train/Reset to run the perceptron

• Attempt XOR to test the impossible

You’ll learn

• How a perceptron separates data with a line

• Why linear separability matters

• The XOR problem and its historical impact

• How hype turned to collapse, setting the stage for multilayer networks