A computer is playing Doom.

Normally, that wouldn't be news. The 1993 game has been made to run on calculators, cameras and almost anything containing a processor.

This machine is different.

At its center are roughly 200,000 living human neurons growing over a silicon chip.

It is called the CL1, built by Australian biocomputing company Cortical Labs.

The headline practically writes itself: human brain cells are playing Doom.

But that isn't quite what is happening.

The neurons don't see monsters or understand the game. Information from the digital environment is converted into electrical stimulation. The cells respond. Electrodes record those responses and software converts them into actions. Feedback changes the network's behavior.

In other words, this is not a miniature person trapped inside a computer.

It is something potentially more useful: a living adaptive processor connected to software.

Cortical Labs has been working on this idea for years.

Its earlier DishBrain system connected cultured neurons to a simplified version of Pong. In research published in Neuron in 2022, neural cultures changed their activity under closed-loop feedback and improved their performance.

CL1 takes that laboratory experiment and packages it into a self-contained machine, complete with electrodes, software and life support for the cells.

Now Cortical has gone another step.

Researchers can access biological computers remotely through Cortical Cloud.

That development may matter more than Doom.

Technologies become industries when people outside the original laboratory can use them.

Cortical has partnered with data-center operator DayOne and the National University of Singapore on an experimental biological-computing deployment.

The first Singapore installation is modest: a rack of 20 systems at NUS Medicine.

But the ambition is unusual.

They want to learn whether biological processors can eventually operate alongside conventional computing infrastructure while consuming dramatically less energy.

That claim remains unproven.

Neurons themselves use remarkably little electrical power, but they also need nutrients, temperature regulation, pumps, sterile handling and eventually replacement.

Until those costs are included in a complete benchmark against silicon performing the same task, nobody can responsibly claim that biological computers have solved AI's energy problem.

The most convincing application may not be replacing GPUs at all.

CL1 gives researchers a programmable network of human neurons.

That could become valuable for studying neurological disease and testing drugs.

Instead of asking only whether a compound kills neurons, scientists could observe how it changes the behavior of an entire living neural network.

Patient-derived neurons could eventually make some experiments more relevant to human biology while reducing dependence on animal models.

For Cortical Labs, that may become a much nearer commercial opportunity than competing with Nvidia.

Switzerland's FinalSpark already gives researchers remote access to living neural systems through its Neuroplatform.

Academic researchers behind Brainoware have also demonstrated organoid-based reservoir computing in Nature Electronics.

Meanwhile, conventional chipmakers are pursuing neuromorphic processors that imitate useful properties of brains without using living tissue.

That competition creates the question that will decide wetware's future:

Can biology do something silicon can't do cheaply enough?

If brain-inspired electronic chips eventually achieve similar efficiency and adaptability, maintaining living neurons inside computers becomes difficult to justify.

There is currently no convincing evidence that CL1's neural culture is conscious.

But the absence of consciousness today doesn't settle the issue forever.

Future cultures could become larger, longer-lived and connected to increasingly sophisticated artificial environments.

Scientists do not have a reliable consciousness meter.

That means rules around donor consent, biological welfare, commercialization and system complexity should probably develop before the technology becomes powerful enough to force the question.

Ignore the next spectacular video-game demonstration.

Watch the boring numbers.

Can independent researchers reproduce the results?

Can biological systems beat conventional computers at a useful task?

What is the real energy cost when the entire life-support system is counted?

How reliable are the cultures after months of operation?

And most importantly: do researchers and pharmaceutical companies keep paying to use them?

Those answers will determine whether Cortical Labs has built the beginning of a new computing industry or one of the most fascinating scientific instruments of this decade.

Either outcome would matter.

Because the important story isn't that brain cells can play Doom.

It's that living human neurons are becoming programmable infrastructure.

And for the first time, that infrastructure is leaving the laboratory.

The most realistic dangers are not armies of conscious brains trapped in server racks.

They are more familiar:

A more distant risk is that researchers eventually create systems with morally significant experiences without recognizing them.

That possibility is uncertain. But uncertainty is precisely why governance should develop alongside capability rather than several years behind it.

Cortical Labs has not built a human brain inside a computer.

It has built something more modest and, in its own way, more scientifically interesting: a programmable interface between living human neurons and digital systems.

The evidence shows that cultured neurons can respond to structured information, change their activity through feedback, and participate in simple computational tasks.

The evidence does not yet show that these systems are conscious, generally intelligent, commercially scalable, or more energy-efficient than silicon after total operating costs are counted.

The strongest near-term opportunity is not replacing AI data centers.

It is creating better tools for neuroscience, drug discovery, disease modelling, and the study of learning itself.

The long-term computing potential remains uncertain. Biology offers extraordinary adaptability and efficiency, but it also introduces variability, fragility, maintenance costs, and moral questions that silicon does not carry.

The CL1 deserves neither ridicule nor blind celebration.

It deserves careful attention.

Because the most important thing about this technology is not that neurons can play Doom.

It is that we have begun turning living human neural tissue into programmable infrastructure and we are doing so before society has fully decided what responsibilities come with it.

What is it?
A hybrid system combining living human neurons, electrodes, life-support hardware, and software.

Is it a human brain?
No.

Did it really play Doom?
It participated in a limited closed-loop control task connected to the game.

Is it conscious?
There is no convincing evidence that it is.

Can it replace GPUs?
Not remotely not with present capabilities.

Could it reduce energy use?
Possibly for specialized tasks, but no independent whole-system comparison has proved it.

What is its most promising use?
Neuroscience, neurological drug testing, disease modelling, and research into adaptive learning.

What should we watch?
Independent replication, energy benchmarks, reliability, paying customers, and ethical oversight.