Vivodyne says AI drug discovery has a data problem, and it built a machine to fix it

A biotech startup argues the missing ingredient isn't compute — it's causal biological data from living tissue

Published: 2026-08-22
Category: Quick Take
Sources: TechCrunch

The argument

AI isn't close to curing cancer — and a biotech startup called Vivodyne says the drug-discovery industry has a data problem, and it built a machine to fix it. Its HIVE modular robotic labs grow 20 kinds of human tissue, then autonomously dose and monitor them, generating the kind of causal biological data that today's AI models lack — data that currently mostly comes from animal testing or studies of single cells and proteins, not living tissue.

"Absent human testing, what are these AI models going to do?" asks Andrei Georgescu, Vivodyne's CEO and co-founder. "They're going to cure cancer in mice." Even Anthropic CEO Dario Amodei wrote this weekend that claims AI will cure cancer have become more cliché than credible — "the thing that will work is actually curing cancer."

The reality check

The grand claims are everywhere. Amodei himself has floated them; Sam Altman cites curing cancer as justification for OpenAI's AGI push; DeepMind's Demis Hassabis said last year AI could cure all disease within a decade. But the actual results remain tepid. A handful of AI-designed drugs have reached human trials — one as far as Phase III — yet the roadblocks aren't necessarily ones AI can solve today. Nobel-winning AlphaFold advanced understanding of protein structure but hasn't produced a new drug. Isomorphic Labs expects its first trials by year's end. In February, the company wrote that true drug discovery requires "highly accurate predictive models across an expansive range of biochemical properties."

The fix

Georgescu wants "a sanity check" — existing models lack the data to capture human biology's complexity. It's a problem the pharmaceutical industry already knows: 90% of drugs that pass animal testing for efficacy don't receive regulatory approval for humans. Vivodyne's alternative: tissue models that closely match real human organs. The company says its liver cells have 94% predictive accuracy against human toxicity trials, its airway tissue matches real behavior 96% of the time, and its bone marrow achieved 100% concordance across 20 chemotherapy drugs.

Why it matters

The lesson here is that the bottleneck in AI-for-science isn't model size — it's data quality. Vivodyne's bet is that real, causal human-tissue data, not protein-level abstractions, is what lets models finally cross from laboratory to bedside. If even half its concordance claims hold up, it's addressing the exact failure point that has kept the "AI cures cancer" narrative rhetorical.

Source: TechCrunch, "AI isn't close to curing cancer. This startup says it knows what it will take," Tim Fernholz, August 19, 2026.