The First Computer Made of Human Brain Cells Just Did the Impossible
Part 1, how DishBrain and the CL1 came to be, the Doom demo, and why we couldn't tell if the cells felt anything: https://youtu.be/RsuIysO4e1E
In Melbourne, a box called the CL1 runs a maze on living human brain cells grown from donor skin and blood. A standard learning AI gets the same maze and the same number of moves. The cells get thirty. By the third round they are finding the food, and the silicon agent next to them is still walking into walls.
━━━━━━━━━━━━━━━━━━━━━━━━
📡 THE SPACIALIZE BRIEFING — our free newsletter
The story behind each video, plus what we're investigating next in space. No filler.
👉 https://tinyurl.com/2bslz6ov
━━━━━━━━━━━━━━━━━━━━━━━━
This video follows Cortical Labs' 2026 test step by step: the six-by-six grid world with food and a scent, the static the cells receive after every wrong move, the two-minute rests that mattered more than extra moves, 1,296 encoding settings over 4,000 hours across 26 cultures, and the twelve settings that beat a tuned deep Q network given the same number of interactions. Then the mechanism underneath it, the free energy principle, and why a living network learns from a handful of tries where silicon needs a mountain of examples. Brainoware at Indiana University, the organoid that learned a chaotic map with a fraction of the training. And who paid for that property: In-Q-Tel, the venture arm of the CIA, an Australian national-security research grant, Tianjin's MetaBOC organoid robot, and the twenty-unit rack of sixteen million living neurons switched on at the National University of Singapore in August 2026.
The 4,000-hour result comes from a preprint written by Cortical Labs' own team. Nobody outside the group has reviewed those numbers yet.
SOURCES
Zhou, Tanneberg, Habibollahi, Loeffler et al., "Embodied Neurocomputation: A Framework for Interfacing Biological Neural Cultures with Scaled Task-Driven Validation," arXiv preprint, 13 May 2026
https://arxiv.org/abs/2605.13315
Cortical Labs, CL1
https://corticallabs.com/cl1
Mnih et al., "Human-level control through deep reinforcement learning," Nature, 2015
https://www.nature.com/articles/nature14236
Cai et al. (Feng Guo lab), "Brain organoid reservoir computing for artificial intelligence," Nature Electronics, December 2023
https://www.nature.com/articles/s41928-023-01069-w
MIT Technology Review on Brainoware, 11 December 2023
https://www.technologyreview.com/2023/12/11/1084926/human-brain-cells-chip-organoid-speech-recognition/
Brett Kagan in the Journal of Medical Internet Research, via News-Medical, 29 May 2026
https://www.news-medical.net/news/20260529/Harnessing-human-brain-cells-to-create-biocomputers.aspx
TechCrunch, Cortical Labs' $10M round including In-Q-Tel, 19 April 2023
https://techcrunch.com/2023/04/19/cortical-labs-raises-10m-for-its-pong-playing-stem-cells-which-eventually-could-power-ai
Defence Connect, National Intelligence and Security Discovery Research Grant to Monash and Cortical Labs, 25 July 2023
https://www.defenceconnect.com.au/industry/12440-research-merging-human-brain-cells-with-silicon-chips-receives-national-security-grant
China Daily, Tianjin University and SUSTech brain-on-chip (MetaBOC), 5 July 2024
https://global.chinadaily.com.cn/a/202407/05/WS66879e62a31095c51c50c9b1.html
New Atlas, organoid-controlled robot, 30 June 2024
https://newatlas.com/robotics/brain-organoid-robot/
STAT, "Brain organoid pioneers fear backlash over biocomputing," 17 November 2025
https://www.statnews.com/2025/11/17/brain-organoid-pioneers-fear-backlash-over-biocomputing/
Salem et al., Nature comment on tissue donors and biocomputers, 30 July 2026
https://www.nature.com/articles/d41586-026-02316-8
Interesting Engineering, biological data centre at NUS with DayOne and Cortical Labs, 21 August 2026
https://interestingengineering.com/innovation/world-first-biological-data-center
OUR LINKS
Newsletter: http://tinyurl.com/2bslz6ov
Podcast: http://open.spotify.com/show/033JRf03QckID9kOFuIhYu
Merch: http://spacialize-shop.fourthwall.com
#Biocomputer #CorticalLabs #OrganoidIntelligence