Brain Cells Master Doom: Cortical Labs’ Biological Computer Reaches Major AI Milestone

Biological Computer Brain Cells Just Learned to Play Doom in One Week — And It Changes Everything

Estimated reading time: 12 minutes

Key Takeaways

• Biological computer brain cells — roughly 200,000 human neurons — learned to play Doom in under a week.
• Cortical Labs’ CL1 uses living neurons on a microelectrode array with two-way electrical communication.
• The jump from Pong to Doom represents a roughly 3,000% improvement in learning efficiency.
• Biocomputing could help address the AI energy crisis with massive energy-efficiency advantages.
• Real-world applications include robotic limb control, diagnostics, drug discovery, edge computing, and brain-computer interfaces.
• Investment and infrastructure are scaling, alongside ongoing ethical considerations and scientific limitations.

Introduction: When Living Neurons Meet a Classic Shooter

Biological computer brain cells just did something remarkable. A clump of human neurons — roughly 200,000 of them — learned to play Doom, one of the most complex classic video games ever made. And they did it in under a week.

This is not science fiction. This is happening right now, in 2026, at an Australian biotech company called Cortical Labs.

The achievement is being called a watershed moment in neuromorphic computing. It builds on earlier breakthroughs, signals a massive leap in biocomputing capability, and opens up real-world applications that were considered decades away — things like controlling robotic arms with living neural networks.

Here is everything you need to know about what happened, why it matters, and where this technology is headed.

Section 1: What Is a Biological Computer — And How Does It Work?

Neuromorphic Computing Meets Living Brain Cells

A biological computer is not a machine made of metal and circuits in the traditional sense. It is a system that uses living human neurons — real brain cells — to process information.

Cortical Labs pioneered this concept with a device they call CL1. Here is how it works:

• Human neurons are grown from stem cells in a lab setting.
• Those neurons are placed on top of a microelectrode array — a flat surface covered in tiny sensors.
• The array can both send electrical signals to the neurons and receive signals back from them.
• This two-way communication allows the neurons to receive information (like a game screen) and respond with actions (like moving a character).

Each CL1 unit contains approximately 200,000 human neurons. The device includes integrated life-support systems to keep those neurons alive and functional for up to six months. No external computing infrastructure is required.

The CL1 was officially launched in March 2025 as the world’s first commercially available, code-deployable biological computer. It comes with a cloud-based API and a touchscreen interface. Developers can program it using Python — the same language used to build most modern AI systems.

Each unit costs approximately $35,000.

This is organoid intelligence in practice — the idea, roadmapped by Johns Hopkins researchers in Frontiers in Science, that 3D brain organoid cultures connected to sensors and machine interfaces could form the basis of a new computing paradigm.

“It’s this accessibility and this flexibility that makes it truly exciting.” — Brett Kagan, Cortical Labs

Cortical Labs unveils biological data center prototype

Frontiers in Science: organoid intelligence roadmap

Section 2: From Pong to Doom — A 3000% Leap in Learning Speed

Organoid Intelligence Accelerates: The DishBrain to CL1 Journey

To understand how big the Doom achievement is, you need to know what came before it.

In 2021, Cortical Labs introduced DishBrain — a chip containing over 800,000 living neurons that learned to play Pong. That project took more than 18 months of painstaking scientific work. Researchers had to carefully design training methods, iterate constantly, and work directly with biological systems that most programmers had no experience handling.

Now compare that to Doom:

• The Doom demonstration used only about 200,000 neurons — roughly a quarter of the neurons used for Pong.
• An independent developer named Sean Cole, who had relatively little prior experience working with biology, taught the chip to play Doom using Python.
• He did it in approximately one week.

That is a learning efficiency improvement of roughly 3,000%.

What made this possible? Cortical Labs built a Python-based interface that makes it far easier for developers — not just biologists — to program these biocomputing chips. You no longer need a biology PhD to work with living neural networks.

The Doom gameplay required neurons to:

• Receive real-time visual information encoded as electrical stimuli.
• Send motor commands in response.
• Adapt their behavior based on feedback — a process fundamentally different from static “feed-forward” systems.

Dr. Andrew Adamatzky at the University of the West of England described the leap clearly:

“Doom is vastly more complex than earlier demonstrations, and successfully interacting with it highlights real advances in how living neural systems can be controlled and trained.”

The neurons played Doom better than a randomly firing system. They did not match elite human players — but that is not the point. The point is adaptive, real-time, goal-directed learning in a living neural network. That has never been demonstrated at this scale and speed before.

Professor Steve Furber at the University of Manchester acknowledged the significance while noting that important questions remain — like how neurons “understand” what is expected of them without eyes or prior experience of visual environments.

PopSci: human brain cell computer plays Doom

Cognitive Wonderland: did brain cells on a chip really

Science News: brain cells organoids video game Doom

Cut Costs. Boost Output. Stay Ahead.

If smarter systems can learn faster with less energy, imagine what your business can do with the right automation. Flowgrammer helps you eliminate repetitive work, reduce operational drag, and scale without hiring headaches.

👉 Book Your Free Automation Audit

Section 3: Why Biocomputing Could Solve the AI Energy Crisis

The Power Problem in Neuromorphic Computing

Artificial intelligence has an energy problem. A serious one.

• Training GPT-3 consumed enough energy to power 120 houses for an entire year.
• Training GPT-4 required approximately 50 times more energy than GPT-3.
• The International Energy Agency projects that data centers will consume 3% of all global electricity by 2030.

This is not sustainable. And it is one of the biggest reasons the biocomputing field is drawing serious attention from investors, governments, and enterprise technology buyers.

Here is where biological neural networks change the game:

• Human neurons process information at an estimated 10⁶-fold (one million times) greater energy efficiency than modern machines.
• Even silicon-based neuromorphic chips — like Intel’s Loihi 2 — deliver 100× energy savings compared to conventional processors.
• BrainChip’s Akida Pulsar delivers 500× lower energy consumption than standard AI processing cores.

Now imagine combining the efficiency of neuromorphic architecture with actual living neurons.

Johns Hopkins researcher Thomas Hartung, who coined the term organoid intelligence, framed this directly:

“Organoid intelligence launches a new era of fast, powerful, and efficient biocomputing.”

The neuromorphic computing market reflects this optimism. Analysts project it will reach $37.18 billion by 2034, growing at a CAGR of 19.9%.

The energy efficiency advantage alone makes biocomputing worth watching — even if every other performance metric remained equal to conventional AI. But they do not. Biological systems also show superior sample efficiency, meaning they learn from far less data than deep reinforcement learning algorithms require.

Polaris Market Research: neuromorphic computing market

Future Markets: neuromorphic computing sensing market

arXiv: 2405.16946v1

Section 4: Real-World Applications Beyond Video Games

From Doom to Robotic Arms — Biocomputing Goes Practical

Doom is the headline. But the actual implications stretch far beyond gaming.

Dr. Yoshikatsu Hayashi at the University of Reading put it plainly:

“[Playing Doom] is like a simpler version of controlling a whole arm.”

Hayashi and his team are working toward using biological computers to control robotic limbs — a task that requires exactly the kind of real-time, adaptive decision-making the CL1 demonstrated in the Doom experiment.

Other near-term application areas include:

• Medical diagnostics — biological neural networks that can identify patterns in patient data in ways silicon-based systems cannot.
• Drug discovery and testing — using organoid intelligence systems as more ethical alternatives to animal testing.
• Edge computing — low-power, adaptive systems that can process information locally without relying on energy-hungry cloud infrastructure.
• Brain-computer interfaces — bridging the gap between biological and digital systems in ways that could enhance human capability.

Concurrent research at UC Santa Cruz showed that mouse brain organoids performing a cartpole balancing task were capable of reinforcement learning when provided with electrical stimulation feedback. Researchers noted that long-term memory retention will require more complex neurochemical pathways — including dopamine reward circuits — that have not yet been replicated in vitro.

This is an important constraint. Biocomputing systems still face real scientific challenges. But each experiment narrows the gap.

Dr. Adamatzky summarized the broader potential:

“What’s exciting here is not just that a biological system can play Doom, but that it can cope with complexity, uncertainty, and real-time decision-making. That’s much closer to the kinds of challenges future biological or hybrid computers will need to handle.”

Intelligence in a dish: Hopkins researchers advance biocomputing

Frontiers in Science: 10.3389/fsci.2023.1017235

Section 5: Investment, Infrastructure, and the Race to Scale Biocomputing

Neuromorphic Computing Market Heats Up — Follow the Money

The Doom experiment was not just a scientific demonstration. It was a commercial signal.

Here is what has happened in the biocomputing investment and infrastructure space in parallel:

Cortical Labs’ Data Center Deployments:

• Partnered with DayOne data center operators to install 120 CL1 units in Melbourne, Australia.
• Planning a phased deployment of up to 1,000 units in Singapore.
• An initial validation phase of 20 CL1 units is underway at the National University of Singapore’s Yong Loo Lin School of Medicine.

Recent Funding Activity:

• In March 2026, Cortical Labs secured investment from Malaysia’s sovereign wealth fund Khazanah Nasional, channeled through Gobi Partners’ Dana Impak Ventures.
• Additional backing came from Horizon Ventures.
• Notable participant: Tom Oxley, founder of brain-computer interface company Synchron.

Competitive Landscape:

• The Biological Computing Co. (TBC) raised $25 million in seed funding in February 2026, led by Primary Ventures, to commercialize living-neuron computing systems.

This is not a fringe research area anymore. Sovereign wealth funds, established venture capital firms, and leading academic medical institutions are all placing bets on biological computing infrastructure.

The race is on to determine who builds the first scalable, commercially viable biocomputing data center — and who controls the software layer that sits on top of it.

Gobi Partners invests in Cortical Labs expands biological computing hub

Human brain cells set to power two new data centers thanks to Cortical Labs

Section 6: Ethical Considerations and Scientific Limitations

Biocomputing Ethics — What Happens When Neurons Are Involved?

Whenever human biology is involved in technology, ethical questions follow. Biocomputing is no different.

Key ethical and scientific considerations include:

Are these neurons conscious?

Brett Kagan addressed this directly: “Yes, it’s alive, and yes, it’s biological, but really what it is being used as is a material that can process information in very special ways that we can’t recreate in silicon.” The neurons are not a brain. They are a processing material.

Is it ethical to use human neurons in computing?

Cortical Labs argues that using lab-grown neurons derived from stem cells is ethically superior to animal testing, a common method in neuroscience research. No animals are harmed. The neurons are grown specifically for research and commercial purposes.

What are the current limitations?

• Long-term memory in these systems requires dopamine reward pathways that have not yet been replicated in vitro.
• We still do not fully understand how the neurons “know” what is expected of them or how they interpret visual data without sensory organs.
• Performance in complex tasks remains well below human capability.
• Regulatory frameworks governing the use of human biological material in commercial computing systems are still nascent and evolving.

What about the critics?

Some researchers initially suggested the neurons were simply acting as a static “feed-forward” system — that the real intelligence came from the encoder and policy learning software, not the neurons themselves. However, expert consensus confirms that the neurons themselves undergo adaptation through membrane potential changes and synaptic weight adjustments — genuine biological learning.

These are not simple relay switches. They are adapting.

Cognitive Wonderland: did brain cells on a chip really

arXiv: 2405.16946v1

FAQ: 10 Questions About Biological Computer Brain Cells and the Doom Breakthrough

Conclusion: The Future of Computing May Be Alive

The story of biological computer brain cells learning to play Doom in a week is not just a clever headline. It is a signal.

Computing is changing at a fundamental level. Living neurons can process information in ways silicon cannot replicate. They are more energy efficient, faster at learning from limited data, and capable of real-time adaptive decision-making. The infrastructure is being built. The investment is flowing. The science is accelerating.

Biocomputing will not replace every form of artificial intelligence or conventional computing. But it will carve out a significant role — particularly where energy efficiency, adaptive learning, and real-time decision-making matter most.

The question is no longer whether biological computing is real. It is how fast it will scale — and who will be ready when it does.

Sources:

PopSci: human brain cell computer plays Doom

Tom’s Hardware: human brain cells set to power two new data centers

Data Center Dynamics: Cortical Labs unveils biological data center prototype

JHU Newsletter: intelligence in a dish

TechNode Global: Gobi Partners invests in Cortical Labs

arXiv: 2405.16946v1

Frontiers in Science: 10.3389/fsci.2023.1017235

Polaris Market Research: neuromorphic computing market

Future Markets: neuromorphic computing sensing market

Cognitive Wonderland: did brain cells on a chip really

Science News: brain cells organoids video game Doom