Australian Biotech Cortical Labs Trains Human-Neuron Biocomputer to Navigate the 3D World of DOOM

Cortical Labs, a pioneering biotechnology startup based in Australia, has reached a landmark milestone in the evolution of biological computing. By March 2026, the company successfully demonstrated real-time adaptive learning on its sophisticated CL-1 platform. This breakthrough involved training a system powered by cultured human neurons to navigate the complex three-dimensional environment of the classic video game DOOM. This achievement marks a significant leap forward from the company's 2021 demonstration, where a predecessor system known as DishBrain mastered the simpler two-dimensional mechanics of Pong.

The CL-1 platform is marketed as the world's first programmable biological computer, utilizing approximately 200,000 human neurons derived from blood stem cells. These living cells are integrated onto a High-Density Microelectrode Array (HD-MEA), creating a bridge between biological matter and digital processing. To play DOOM, the system requires a complex translation process where visual data from the game is converted into patterns of electrical stimulation. These pulses are directed at the neural culture, and the resulting cellular responses are interpreted as specific in-game actions, such as movement, identifying enemies, and firing weapons. Chief Scientific Officer Brett Kagan explained that the learning process relies on structured feedback: predictable signals act as rewards for correct actions, while irregular signals serve as penalties, encouraging the biological system to adapt its behavior.

Demonstrating the accessibility of this wetware, independent developer Sean Coe managed to program the system to play DOOM in just one week. He utilized a proprietary operating system equipped with a Python API, highlighting a successful abstraction layer between traditional software code and biological hardware. Despite this rapid integration, Cortical Labs officials remain realistic about the current capabilities of the system. They acknowledge that the neural cultures currently perform at a novice level, lacking the advanced spatial memory or strategic foresight required for high-level gameplay or complex navigation.

Hon Weng Chong, the CEO of Cortical Labs, clarifies that the CL-1 is not intended to replace traditional silicon-based processors in general computing. Instead, the technology is aimed at specialized physical AI applications, such as autonomous drones and robotics, which require real-time responsiveness and extremely low power consumption. This strategic focus addresses the growing global energy crisis within the artificial intelligence sector. The energy efficiency of biocomputing is a major selling point; a single CL-1 unit consumes only about 30 watts of power, which is a fraction of the energy required by a modern high-performance graphics processing unit (GPU).

To commercialize this technology, the Melbourne-based firm has announced the launch of its first Bio-Data Centre prototype, which houses 120 CL-1 units. Furthermore, Cortical Labs is expanding its international footprint through a partnership with the firm DayOne to establish a much larger facility in Singapore. This upcoming venture is expected to scale up to 1,000 CL-1 units across several development phases. Developed in collaboration with the Yong Loo Lin School of Medicine at the National University of Singapore, this facility will be the first of its kind outside of Australia, serving as a critical testing ground for the scalability of wet computing architectures.

Since its founding in 2019, Cortical Labs has been at the center of both scientific excitement and ethical scrutiny. The company's 2022 demonstration of the DishBrain system sparked intense discussions regarding the moral implications of using living human cells for computational tasks. To foster further research and development, the company now offers the Cortical Cloud, a service that provides researchers worldwide with remote access to its live neural networks. This cloud-based approach allows scientists to experiment with biological intelligence without the need for their own specialized laboratory infrastructure.