Some Scientists Have ‘Magic Hands’ in the Lab. This A.I. Is Learning Why.

“On a recent afternoon, scientists buzzed around a lab in Cambridge, Mass., performing experiments. Their equipment was standard: a lab hood for working with dangerous chemicals, an incubator for growing cells. Their procedures were identical to those in biology labs everywhere.

 

A closer look revealed some oddities, though. Along with lab coats and gloves, most of the scientists wore miniature cameras on headbands. Three additional cameras peered down at each work station from a shelf.

 

Lab notebooks were strangely absent. Instead, the scientists quietly narrated their work, murmuring into microphones. A team huddled at one end of the lab, inspecting videos of the experiments.

 

This was the real research taking place in this lab. With a system of sensors and software intended to capture science as it happens, down to the millisecond, scientists were training artificial intelligence models to recognize every object the scientists used and every action they performed.

 

The A.I. even composed its own narrative, describing every few seconds of video with sentences like, “The operator resuspends the pellet by pipetting it up and down 10 times.”

 

The technology is the creation of a start-up called Transfyr, which came out of stealth mode this week with $25 million in seed funding. The company is tackling an age-old challenge in science: the hidden factors that make some experiments succeed and others fail.

 

Failure can take many forms. Researchers may spend months preparing a line of engineered cells, only to have them mysteriously die along the way. A government Covid test seems to work in one lab, but fails to detect the virus when others use it.

 

A biotech company creates a promising new drug; when it hands off the protocol for large-scale production, suddenly it just doesn’t work.

 

“This is just an incredibly painful problem,” said Anna Marie Wagner, a co-founder of Transfyr. “There’s finger-pointing back and forth. Was your protocol wrong? Or did you screw something up? These are very, very expensive mistakes in terms of time, money, and lives.”

 

Success can be just as mysterious as failure. Some researchers consistently get experiments to work, earning befuddled admiration from colleagues. Scientists even have a special term for this gift: magic hands.

 

The idea may come as a surprise to people who don’t spend their lives in labs. Science is not supposed to be magic.

 

When scientists carry out experiments, they keep careful records, both in lab notebooks and later in published scientific papers. Other researchers use that information to repeat the experiment.

 

But every scientist discovers sooner or later that essential knowledge is not necessarily written down.

 

“A protocol is a recipe,” said Jonathan Livny, a senior research scientist at the Broad Institute who has collaborated with Transfyr. “You can give somebody a recipe, and it’s not going to make them a great chef.”

 

Like apprentice chefs, scientists spend years in training, shadowing experts, asking questions and trying out procedures for themselves.

 

To make an experiment work, they may have to make thousands of minor decisions over the course of a day. A tube needs to be shaken — let it whir on a vibrating platform, or just flick it back and forth by hand?

 

“The really good people sometimes don’t know why they’re that good,” Dr. Livny said. “They don’t remember the 20 times they did something another way and it failed. They’ve blocked all that sadness out.”

 

In the 1960s, the Hungarian chemist and philosopher Michael Polyani gave this mysterious expertise a name: tacit knowledge. “We know more than we can tell,” he liked to say.

 

Scientists have generally come to agree with Polanyi that tacit knowledge is an essential part of research. But it can also slow down progress.

 

In theory, science moves forward as researchers build on previous work. When a new study comes out, other scientists will attempt to replicate the work. If the original result was correct, a replication ought to produce the same results.

 

Surprisingly often, it doesn’t. And tacit knowledge is part of the problem. Scientists can’t know exactly how to replicate a study simply by reading about it.

 

Renee Wegrzyn, a co-founder of Transfyr, became keenly aware of this gap while doing research as a postdoctoral researcher. She jotted down records of her experiments on proteins in a lab notebook, but it captured only a fraction of her efforts.

 

When Dr. Wegrzyn published her results, she had to distill her records even further. “I did a postdoc that was three years long, and then it got summed up in a three-page paper,” she said.

 

Missing from those notes and reports were all the little decisions Dr. Wegrzyn had made while performing her research.

 

“All of those details could matter, but we just don’t know if they matter,” said Brian Nosek, an expert on replication at the University of Virginia who is not involved in Transfyr.

 

When a replicated study fails to produce the original results, that may mean the original study was wrong. Or maybe it was right, and the replicating scientists made an unwitting blunder. It can be hard for teams of scientists to agree on what happened.

 

In 1993, for example, the psychologist Frances Rauscher and her colleagues reported that people who listen to Mozart get a temporary boost on reasoning tests. The buzz around the so-called Mozart effect led other researchers to run experiments of their own.

 

Many of them couldn’t find any significant benefit, but Dr. Rauscher brushed off these failures. The other scientists didn’t find the Mozart effect because they weren’t good enough at running experiments, she said — they were not good scientific chefs.

 

The search for the Mozart effect went on for years. In the end, other scientists found little evidence that it exists.

 

Conflicts like these led some scientists to search for new ways to communicate tacit knowledge. In 2006, the biologist Moshe Pritsker created a journal called JoVE, which published videos of scientists carrying out complex experiments.

 

In 2014, the biologist Lenny Teytelman created a place where scientists could share their protocols and talk about them. Researchers use the website, called Protocols.io, to help train new members of their labs when the seasoned scientists with magic hands have moved on.

 

With Transfyr, Dr. Wegrzyn and Ms. Wagner are building a system that can slurp up enormous amounts of data about what happens in a lab, in the form of video, audio and sensor logs from lab equipment. Transfyr even tracks the source of supplies, down to the lot numbers of glove boxes.

 

(Why? “Maybe there’s a fume coming off the glove that is changing your experiment,” Ms. Wagner said.)

 

Dr. Nosek said that this new approach might reveal some secrets about why experiments succeed or fail. “What I really like about them is they’re going gangbusters into trying to unpack every detail,” he said. “It makes a lot of sense to go all in, and then figure out what actually matters.”

 

The company feeds its digital record of experiments to computers that have been trained to recognize the equipment used in biochemistry experiments. Artificial intelligence systems track the equipment, along with the hands of scientists, and decipher each action.

 

This analysis has revealed that lab workers are performing the same experiment in many different ways. “The variation we see even among well-trained scientists is pretty jaw-dropping,” Ms. Wagner said.

 

Sometimes those variations matter.

 

Recently, Transfyr hosted technicians from Dr. Linvy’s lab, to watch them extract RNA from cells. Charlie Loy, famous for her magic hands, was unwittingly doing a key step wrong, letting chemicals react for 120 seconds instead of 90 seconds.

 

“I didn’t realize that was even happening,” Ms. Loy said. The mistake happened because she turned her timer on after starting the chemical reaction, not before. But it turns out that the extra time from that mistake improved the experiment.

 

In another trial, Transfyr has found that a protocol that takes six hours for one researcher may take eight hours for another. “It’s not a hard thing to see,” Ms. Wagner said. “But there’s no way to capture it at scale manually.”

 

Transfyr has signed up a diagnostics company and other clients who want the software to track their research. Ms. Wagner said that ultimately, it might be possible for the system not just to uncover mistakes but to document unexpected breakthroughs.

 

Rather than sifting notebooks for clues, the system can look back at every second of the work while querying A.I. about unusual things the workers did without realizing it.

 

Dr. Nosek said that Transfyr would have to present detailed results to prove that this system really does make science better. “Take 50 labs, track half the experiments with this stuff and don’t track the other half — then look for differences in progress,” he suggested.

 

Harry Collins, a sociologist at Cardiff University who has studied tacit knowledge for over 50 years, said he was optimistic about the effort because it was focused on molecular biology — a field that has matured to the point that many of its fundamental mysteries have been worked out.

 

But he was skeptical that Transfyr’s approach would reveal much in fields where scientists agree on far less, even about which experiments will reveal something important.

 

“You might be lucky now and again, but it’s not a silver bullet for scientific problems,” Dr. Collins said.” [1]

 

1. Some Scientists Have ‘Magic Hands’ in the Lab. This A.I. Is Learning Why. Zimmer, Carl.  New York Times (Online) New York Times Company. Aug 27, 2026.