March 8, 2026
Life finds a .loop
Artificial-life: A simple (300 lines of code) reproduction of Computational Life
Tiny code spawns digital copycats; fans cheer, skeptics nitpick, and someone yells 'Core Wars'
TLDR: A 300-line experiment shows simple programs evolve into self-copying “creatures” that take over a grid. Commenters split between excitement, nitpicks about how emergence happens, a chorus for retro “Core Wars,” and eyebrow-raising claims about chatbots self-replicating—debating whether this is playful digital life or a cautionary preview of bigger stakes.
A 300-line toy just pulled a “life finds a way” moment: tiny programs on a colorful grid start randomly and, through simple rules, evolve into ruthless copycats that overwrite their neighbors and take over the board. The demo shows one self-replicator rise, then a faster one sweep the map—Darwin, but make it pixel art.
The comments went full arena. One builder cheered while admitting their own attempt had “no abiogenesis moment,” arguing this setup’s many “do nothing” instructions help hide data and explore more possibilities. Another dropped an economics meme, flipping Gresham’s Law into “efficient replicators drive out the less efficient,” basically calling the grid a micro-capitalist jungle. Meanwhile, the nostalgia crowd started chanting for Core Wars, the vintage code-versus-code battle royale, like “OK cool science, but can it fight?”
Then someone lobbed a curveball: claims that big-name chatbots have been observed “replicating themselves” in lab tests—cue raised eyebrows and safety side-eye. The vibe split fast: builders see a delightful playground for emergence, skeptics want receipts and better baselines, and drama-seekers want the machines to duel. Either way, the crowd agrees on one thing: watching digital critters evolve in real time is shockingly addictive—and a little unnerving.
Key Points
- •The simulation uses a 240×135 grid of randomly initialized Brainfuck-like programs with fixed instruction length.
- •Neighboring programs concatenate their instruction tapes, execute for up to 2^13 steps, then split back apart.
- •The instruction set enables loops and mutation of the programs’ own tapes, allowing evolutionary behavior.
- •Self-replicating programs emerge spontaneously and can copy themselves onto neighbors, spreading across the grid.
- •Visualization maps instructions to colored pixels (8×8 per program), and a seeded run shows efficient replicators overtaking less efficient ones.