My first (of hopefully many) publication at @CorticalLabs. Thrilled to share our ideas about the ethics of investigating morally relevant states in embodied neural systems (like our neural cells), with @ANeuroExplorer, @LomaxBoyd and @juliansavulescu
https://t.co/D6PwujxySX
CORTICAL LABS PRESENTS: 🧠💾 CORTICAL DOS! 💾🧠 https://t.co/eDoTNgktxX
1993's most revolutionary program to help you understand your synthetic biological intelligence experiments, built here in Melbourne (in 2026)!
@SterlingCooley Never seen this, pretty cool! I do agree that it's wild to see people who eat meat without batting an eyelid turn around and freaking out about less neurons than in the brain of a cockroach 😅
I'm incredibly proud to announce the launch of the Cortical Cloud today and the establishment of the world's first Biological Data Centre here in Melbourne with 120 Internet connected CL1 units for researchers and developers to access anywhere in the world.
To add to this announcement is our plan to a larger 1000 CL1 Biological Data Centre facility with DayOne as a partner in Singapore.
My hope is that technology will unlock new capabilities for researchers and developers using Biological Computing for intelligence and biomedical research using less energy and data than our regular AI/ML approaches.
Try it out now at https://t.co/iPKMqhTd2k and https://t.co/s0t1u0D5m2 for the news announcement on Bloomberg.
🎯 BIOLOGICAL COMPUTING ECONOMICS: DOOM-PLAYING NEURONS
Lab-grown human brain cells just beat GPT-4 at playing Doom.
Not a typo. Cortical Labs built a $35K computer using actual human neurons. They trained it to play Doom in 7 days.
Here's why the economics matter:
💰 Hardware costs:
• CL1 bio-computer: $35K
• Traditional GPU cluster: $50M
• 1,429× cheaper
⚡ Training efficiency:
• Doom training: 7 days (biological)
• Pong training: 365 days (same system, earlier attempt)
• GPT-4 training: ~$100M + months
• 52× faster learning curve
🔋 Power consumption:
• CL1 rack (30 units): 1 kW
• GPU rack: 30 kW
• 30× more efficient
📈 Market opportunity:
• Pharma AI market: $1.94B (2025) → $16.49B (2034)
• Applications: Drug testing, disease modeling, neural research
• Cloud rental model: Developers worldwide can access CL1
Why this changes everything:
Silicon AI learns through brute-force computation. Biological intelligence adapts organically.
The CL1 outperforms GPT-4 in speed and latency. Not because it has more compute. Because it has real neurons that learn differently.
Pharmaceutical companies pay billions for drug testing. Bio-computers can model human brain responses without human trials.
The economics of hybrid intelligence:
We just crossed the threshold where biology beats silicon on cost, speed, and efficiency.
Cortical Labs shipped 115 units in 2025. Cloud rental model launching. GitHub API available.
This is past sci-fi. It's a $35K product you can buy today.
The future of computing is biological.
My first (of hopefully many) publication at @CorticalLabs. Thrilled to share our ideas about the ethics of investigating morally relevant states in embodied neural systems (like our neural cells), with @ANeuroExplorer, @LomaxBoyd and @juliansavulescu
https://t.co/D6PwujxySX
We’ve been a little quiet at @CorticalLabs for a while as we cook up some delicious APIs for developers to work with the CL1 both on device and on the cloud. Here’s what someone did with 10 days of access to a CL1 via the Cortical Cloud announcement - https://t.co/IqGkMn9o4G
Brain cells in a dish playing...DOOM?
A recent graduate, Sean, used our CL API to get living neural cultures to play Doom.
And just in a few days...
Biological neural systems as an information processor has moved off the lab-bench and onto your laptop.
https://t.co/MtHL5TirhB
Will we ever understand how consciousness arises and what would it mean if we did?
Our new paper proposes an iterative process for investigating morally relevant states, ranging from single cells to intact brains.
Check it out here: https://t.co/ZYd3NFkw0h
New Active Inference videos this week:
@Shamburgularara has recorded MorphStream #002 and this will premiere in a few hours:
ActInf MorphStream 002.1
Alon Loeffler @alonloeffler@CorticalLabs
"Biological-Inspired Intelligence: From Neuromorphic Nanowire Networks to Neurons" https://t.co/8imE6qLfll
Then @DariusPW has recorded Insights #005 and this will premiere on Thursday:
Marilyn Stendera @MarilynStendera
Active Inference Insights 005 ~ Autopoiesis, Heidegger, Futurity
https://t.co/WxrobVN8OC
A difference in company philosophy:
@neuralink: put wires on brain.
@CorticalLabs: grow brain on wires.
Cortical Labs just completely changed my dreams and nightmares.
Here it is in Hon Weng’s hotel room. He is showing this off tomorrow at a brain conference in San Francisco.
You get a sneak peak tonight.
@patrickmineault@leafs_s We saw that nanowires could also remember 7 different patterns in working memory, which we linked to this original task.
We don't really make any strong statements about this, but note that the similarities are intriguing :) 2/2
@patrickmineault@leafs_s Hi Patrick!
I'm the author of this article. Hopefully I can answer this!
The original results from the n-back task used in Psychology showed that humans can store about 7 different items in memory. Our study implemented a similar type of task in nanowires. 1/2