A San Francisco startup is using living neurons grown on electrode arrays to make AI-generated video sharper, cheaper, and more coherent
A startup founded by two neurosurgeons just made the jump from science fiction to Amazon Web Services. The Biological Computing Company, known as TBC, is bringing its neuron-powered AI optimization tools to AWS, marking one of the most unusual cloud computing partnerships in recent memory.
The core idea: grow living rat neurons (and human stem-cell-derived neurons) on electrode arrays, then use the biological signals to improve how AI models generate video.
What TBC actually built #
TBC cultivates neurons on multi-electrode arrays containing up to 4,096 electrodes per array. These arrays record how biological neural networks respond to encoded data like images and video, and TBC’s software translates those responses into adapters that can be layered onto existing transformer-based AI models.
The key selling point is that end users never need to touch any biological hardware. TBC’s adapters function as software plugins that slot into models developers are already running.
On July 30, TBC released a proof-of-concept targeting Decart’s OASIS 500M model, a relatively compact video generation system. The results were striking: a 2x improvement in video quality, 4.4x lower inference costs, and more than 3x improvement in coherent video output compared to the base model.
From stealth to AWS in six months #
TBC was founded by Alex Ksendzovsky and Jon Pomeraniec, both neurosurgeons. The company emerged from stealth in February 2026, simultaneously announcing a $25 million seed round led by Primary Ventures and opening its flagship wet lab in San Francisco’s Mission Bay neighborhood.
AI, tech, and the markets they move—in one daily briefing.
Daily. Free. Join 34,000+ readers across crypto, finance, and policy.
The AWS partnership has included lab tours and public demonstrations for AWS clients. TBC went further on September 30, hosting a live demonstration on Twitch where attendees could watch neuron-encoded data being processed in real time.
The company has stated plans to deploy its technology commercially for computer vision and generative video applications before the end of 2026, with a commercial text-to-video model on the roadmap.
Why biological computing, why now #
Generative AI has a well-documented set of problems that pure silicon hasn’t solved elegantly. Object permanence, the ability to keep track of where things are across video frames, remains a persistent weakness in AI-generated video. TBC’s thesis is that encoding visual data through biological neurons captures some of this innate spatial reasoning and injects it back into transformer models as a learned optimization.
Energy efficiency is the other angle. The human brain runs on roughly 20 watts of power. Training a large AI model can consume megawatts.
The competitive landscape and what to watch #
The $25 million seed round is modest by AI startup standards. The real test will be the commercial text-to-video model expected later this year, which will need to demonstrate that its biological optimization holds up across different model architectures and at commercial scale.
Disclosure: This article was edited by Editorial Team. For more information on how we create and review content, see our