Conway's Game of Life

A 2D grid of cells, each black or white. Each step, every cell looks at its 8 neighbors and follows two rules: "alive with 2 or 3 neighbors stays alive; dead with exactly 3 neighbors becomes alive". Everything else dies. From this minimal recipe — Conway 1970 — emerge gliders, oscillators, replicators, and computers powerful enough to simulate themselves.

Controls

speed

Pattern library

Stats

generation0
population0
why Life is the canonical emergence
Conway's rules are absurd in their simplicity:
• survive if you have 2 or 3 living neighbors
• be born if you have exactly 3
• otherwise die

From these two clauses emerge: still lifes (block, beehive, loaf), oscillators (blinker, pulsar), spaceships that crawl across the grid (the glider; LWSS; HWSS), guns that produce a stream of spaceships forever (the Gosper glider gun), eaters that consume incoming spaceships, and ultimately — logic gates, memory, clocks, and finally universal computation.

Paul Rendell built a Turing machine in Life (2002). Adam Goucher built a programmable computer in Life (2010). Nicolas Loizeau built a Life pattern that simulates Conway's Game of Life inside itself. The implication is breathtaking: any program, including this entire web page, could in principle run on a sufficiently large Game of Life board. Life is Turing-complete with a 2-state, 8-neighborhood, deterministic local rule. It demonstrates that computational universality is generic, not exotic.

Click Acorn — a tiny seed of 7 cells. Watch it sputter, then explode into a 1,057-cell pattern over 5,206 generations before settling. The unpredictability isn't randomness; it's the depth of the consequence chain. The rules are deterministic, but the long-term behavior is effectively unknowable except by running it.