Lab-Grown Brains, AI Hype, and How to Read Capability Claims

Every few months a science story arrives announcing that artificial intelligence has been dethroned. This one is about human brain cells grown in dishes, and it is genuinely good reporting. It is also a clean case study in a skill every business owner needs right now: separating what a technology does today from what its promoters need you to believe about tomorrow.

“For now, it seems that in the race toward conscious machines, there will be two lanes. One will be paved with rare earth minerals and silicon—forced, at enormous financial and energy cost, to model the brain from the top-down, in the form of artificial neural networks. Although there’s a lot these systems can do well, they will never be brains. Nothing forged from silicon ever could.”

— Claire L. Evans, WIRED

Our take

Start with the detail the story buries in its final third: one of the field’s leading advocates admits the branding is the point. Thomas Hartung of Johns Hopkins, who helped christen the field “organoid intelligence” in a 2023 declaration, tells WIRED that OI is “above all, good PR—a savvily branded attractor for funding and talent in a wide-open field.” He then says his own slice of the pie is drug development. That is an unusually honest thing for a researcher to say out loud, and it should change how you read the headline above it.

Now look at the hardware. Cortical Labs sells the CL1, a biological computer running roughly 800,000 human neurons, for about $35,000 per unit or $20,000 each in 30-unit racks, with cloud access around $300 a week, according to IEEE Spectrum. The cells stay viable for about six months. So a full rack holds around 24 million neurons. A human brain has roughly 86 billion. That rack is about three hundredths of one percent of a brain, and it needs replacing twice a year.

Karl Friston, the theoretical neuroscientist whose work underpins the famous Pong experiment, is blunt about what this actually is. “The real gift of this technology is not to computer science,” he told IEEE Spectrum. “Rather, it’s an enabling technology that allows scientists to perform experiments on a little synthetic brain.” The person with the largest theoretical stake in the outcome calls it a lab instrument, not a computer.

The commercially real event here has nothing to do with computing. In July 2025 the National Institutes of Health moved to prioritize human-based research over studies relying exclusively on animal models. In September 2025 it committed $87 million to a Standardized Organoid Modeling Center. Two months earlier the FDA had published a plan to phase out animal testing requirements for monoclonal antibodies and other drugs. That is a procurement shift inside a regulated industry with a documented failure problem. Organoids are winning a contract, not a computing race.

Which brings us to the pattern worth stealing. In emerging technology, the capability narrative and the revenue narrative almost always point in different directions. The capability narrative — “AI is dead, organoids are alive” — attracts attention, talent, and funding. The revenue narrative is a boring, regulated, adjacent market where the technology is simply better than the incumbent option. Hartung said this outright to a reporter, and it still ran under a headline about outthinking neural networks.

You will recognize the shape in your own inbox. “Agentic AI” is the organoid intelligence of marketing technology: a compelling label attached to capability that mostly is not in production yet, sold next to plumbing that genuinely works. Data cleanup, deliverability, tagging, routing, transcription, first-draft copy. The plumbing is where agencies and vendors actually earn revenue. The label is what gets the meeting booked.

Here is where we part company with the piece. Its closing bet — “I’m betting on the neurons” — quietly swaps a substrate argument for a schedule. Grant the substrate claim entirely. Maybe silicon can never be a brain. Nothing in the reporting suggests organoids will perform commercially useful general computation on any timeline a business could plan against. Six-month cell viability, a roughly five-millimeter size ceiling caused by necrotic cores, and a per-rack neuron count in the tens of millions are not the specifications of near-term infrastructure. Being right about a destination tells you nothing about the arrival time, and businesses buy arrival times.

The practical test is simple. When a capability claim lands, ignore the demo and look for the procurement change. Federal agencies rewriting funding and approval rules is a signal. Neurons playing Pong is a photograph.

What this means for your business

1. Track procurement, not demos. A demo shows what a technology can be made to do once, under supervision. A funding rule change shows what institutions are willing to pay for repeatedly. When you evaluate an AI vendor, ask who is already paying for this in production and for what specific line item.
2. Treat a rebrand as a warning, not a milestone. Hartung called organoid intelligence good PR for attracting money. Apply that same test to “agentic,” “autonomous,” and “AGI-ready” in martech pitches. Ask what the product did before it got the new name, and what changed besides the name.
3. Buy the capability that already works, at today’s price. Current-generation AI is reliably good at drafting, summarizing, classifying, and routing. Those are real margin gains. Do not pay a premium today for a substrate shift promised for later.
4. If the ROI case depends on capability that does not exist yet, it is R&D. Fund it from a different budget line, with a different success threshold and a named owner. Do not let speculative spend hide inside your marketing budget, where it will quietly crowd out things that work.
5. Set a re-evaluation date instead of placing a bet. Put a calendar reminder twelve months out to recheck the category. Cheap, honest, and it stops you from either overcommitting now or dismissing something that becomes real later.

Claire L. Evans spent months inside these labs and the full piece is worth your time. Read the original reporting at WIRED.

---