AINEWS
Michael Polansky's Mill Valley AI and biology shop says it has held human tissue viable for over a month, the training data for its model, but no outside scientist has reviewed the work.
The next training set isn't a corpus. It's a petri dish.
Outer Biosciences, a small AI-and-biology company in Mill Valley, California, says it has spent years on a system that keeps living human tissue viable outside the body for more than a month. That tissue is the company's training data: alive, capturable, and run through an AI model that the lab's founder, Michael Polansky, has so far only shown to people inside the building.
The company's central claim, on the record in a TechCrunch profile of Polansky published Friday, is that long-lived ex vivo tissue (skin kept alive outside the body, in a dish) is the missing substrate for the next generation of biological AI. Almost no outside scientist has seen the data, the protocol, or the survival curves. The claim is on the record; the work is not.
Polansky is not a biologist by training. He spent three years at the Bridgewater Associates hedge fund before moving to Founders Fund, where he became a principal under the firm's four original partners: Peter Thiel, Sean Parker, Luke Nosek, and Ken Howery. He later ran Sean Parker's family office and helped stand up the Parker Institute for Cancer Immunotherapy, where he remains executive director. He is also the creative, business, and romantic partner of Stefani Germanotta, professionally known as Lady Gaga, a biographical fact that the lab's press footprint leans on but that the work itself does not depend on.
Outer Biosciences fits a recognizable operator type: small, deliberately quiet, with older-money pedigree, run by someone who came up through Founders Fund and Parker-adjacent philanthropy rather than a university biology department. The lab sits in Mill Valley, about 15 miles north of San Francisco, and the company's work has been undisclosed outside its own walls for years.
The technical premise, as Polansky describes it, is that keeping human tissue alive outside the body is hard. Burn-care and dermatology labs have kept skin in culture since the 1980s, but most protocols give out within days or weeks. Stretch that window to a month or more, and the tissue can be instrumented: sensors reading gene expression, morphology, response to perturbations. The stream of recordings becomes the training set for an AI model, and the model learns from biology that is still alive, not from fixed slides or sequenced reads that are static the moment they leave the incubator.
Keeping skin in culture is not new. What Outer is claiming is not the tissue itself but the loop: a sensor-and-AI pipeline that turns living skin into a long-running data source. The novelty lives in the loop, not the dish. Whether the loop is real (whether the data is what Polansky says it is, whether the model is learning anything useful from it, whether a third-party lab could replicate the survival window) is the question the press has not been able to answer, because almost no one outside the company has seen the work.
The watch items are narrow: who else can keep human tissue alive for a month, who else has built the sensor loop on top, and which of these quiet labs lets an outside lab in first. The answer to any of those will tell the reader more about the field than another closed-door profile will.