Cortical Labs Trains 200,000 Human Neurons to Play Doom, Launching Biological Computer
Updated
Updated · sebastianbarros.substack.com · Jun 1
Cortical Labs Trains 200,000 Human Neurons to Play Doom, Launching Biological Computer
14 articles · Updated · sebastianbarros.substack.com · Jun 1
Cortical Labs said its CL1 system used 200,000 lab-grown human neurons to play Doom, which it describes as the first code-deployable biological computer.
The company is pitching the platform as an answer to AI’s power and cooling bottlenecks, arguing living neurons could deliver far better energy efficiency than GPU-heavy data centers.
Cortical Cloud lets developers access the wetware remotely through a browser and run standard Python code on the biological system rather than in a sterile lab.
The neurons had previously been trained to play Pong, and the Doom milestone extends Cortical Labs’ effort to turn neural cultures into a commercial biological computing platform.
If lab-grown brains can learn, where is the line between a tool and a conscious being?
Is this the dawn of hyper-efficient biocomputing or an ethical minefield we are unprepared for?
Human Neurons Trained to Play Doom: The Rise, Technology, and Future of Biohybrid Computing in 2026
Overview
In early 2026, Cortical Labs made a breakthrough by training about 200,000 living human brain cells on a silicon chip (the CL1 system) to play the video game Doom. At first, these neurons struggled with basic actions like moving and shooting, but over time, they learned to target enemies more accurately, showing clear signs of learning. This real-time, goal-directed learning in a biological system highlights the huge potential of biohybrid computing, moving beyond simple games and demonstrating that living neurons can adapt and interact with complex digital environments.