TL;DR: Australian biotech company Cortical Labs has demonstrated its CL1 biological computer — powered by 200,000 living human neurons on a microchip — playing Doom. It’s not beating any records, but the neurons are learning, marking a major step in biological computing.
How Cortical Labs Taught Living Neurons to Play a Video Game
Cortical Labs has done something that sounds like science fiction: trained real human brain cells grown on a microchip to play Doom. The Australian biotech company demonstrated the feat using its CL1 biological computer, a self-contained unit housing around 200,000 living human neurons on a multi-electrode array — a silicon chip that both sends and receives electrical signals from the neural culture.
This isn’t the company’s first foray into neuron-powered gaming. In 2022, Cortical Labs made headlines when its earlier DishBrain system taught lab-grown neurons to play Pong in just five minutes — faster than a standard deep reinforcement learning algorithm would manage in 90 minutes. Doom, however, is a vastly more complex challenge, introducing 3D environments, enemy encounters, and real-time spatial awareness that Pong simply doesn’t require.
Translating Doom Into the Language of Neurons
The core challenge wasn’t just getting neurons to respond — it was teaching them what to respond to. As the Cortical Labs team explained in their February 25 demo video, the solution required converting the digital world of Doom into electrical signals the neurons could actually process.
The system maps gameplay visuals into stimulation patterns delivered across different regions of the neural culture. When an enemy appears on the left side of the screen, electrodes on the corresponding side of the array fire. The neurons then respond with their own electrical patterns, which are decoded into in-game actions — shooting, moving, or turning. Different firing patterns trigger different commands, giving the cells a functional role in controlling the game.
Independent researcher Sean Cole built the first working version of the integration using the Cortical Labs API, running on a CL1 via the Cortical Cloud. The source code has been made publicly available on GitHub.
Not an Esports Champion — Yet
Performance is, understandably, limited. The neurons currently play at roughly beginner level, dying frequently and struggling with the complexity of combat encounters. As Cortical Labs researcher Dr. Brett Kagan put it: “Is it an esports champion? Absolutely not. Right now, the cells play a lot like a beginner who’s never seen a computer.”
That said, the team stresses this isn’t the point. What matters is that the neurons are demonstrably learning and adapting through feedback — and that the company has, in their own words, “solved the interface problem” of communicating in real time between silicon hardware and living biology. The next phase involves refining how information is encoded, how rewards are delivered, and how neural responses are interpreted to push performance further.
It’s a fascinating contrast to how traditional games push hardware to its limits. While games like those arriving on the Nintendo Switch 2 this year rely on conventional silicon getting faster and more powerful, Cortical Labs is betting on a completely different paradigm — biology itself as the computational substrate.
What Is the CL1?
The CL1 is Cortical Labs’ first commercial product, with the first 115 units shipping in 2025. Each unit is a self-contained desktop device with an internal life-support system that keeps neurons alive for up to six months through temperature control, waste filtration, and gas circulation. It runs on a biological operating system called biOS, which lets developers deploy code directly to the living neural network using Python via the CL API.
The neurons themselves are derived from skin or blood samples of adult donors, reprogrammed into induced pluripotent stem cells and then differentiated into brain cells. A full unit retails at $35,000, dropping to $20,000 per unit in 30-unit server rack configurations — which consume only 850 to 1,000 watts, comparable to a single midrange GPU server.
Cortical Labs labels its approach “Synthetic Biological Intelligence” — a deliberate distinction from AI, emphasizing that these are real neurons performing real computation on real silicon, not simulated neural networks.
Doom Has Been Ported to Everything — Including Gut Bacteria
Doom’s legacy as a porting benchmark stretches back decades. The 1993 id Software classic has been run on calculators, printers, ATMs, and even a pregnancy test. Researchers previously demonstrated a version of Doom simulated using gut bacteria, though with significant caveats around what “playing” actually meant in that context. Living human neurons on a biological computer represent arguably the most biologically complex platform Doom has ever run on.
For gaming fans tracking the latest in interactive entertainment — from Bungie’s Marathon server infrastructure updates to new Pokemon Winds and Waves starter reveals — the idea of neurons as future gaming hardware might still sound far-fetched. But Cortical Labs is methodically closing the gap between biological research and practical computing, one demon-infested corridor at a time.
