Pattern Computer builds an engine to find the patterns hiding in data too complex for humans, and too subtle for ordinary machines.
Drive to the end of the Washington map, take a ferry past the orcas, and you arrive in Friday Harbor - a town better known for whale-watching than for computing. Somewhere here sits Pattern Computer, a company of roughly twenty-five people quietly insisting it can find things in data that nobody else can. The claim sounds outlandish until you read the small print: they are not trying to recognize patterns. They are trying to discover them.
The distinction matters more than it looks. Most modern AI is a brilliant matcher - show it ten thousand cats and it will spot the next cat. Pattern Computer is after the cat nobody has named yet: the relationship buried in dozens of dimensions that no human eye, and no conventional algorithm, would think to connect. They first pointed that capability at medicine, on the theory that if you can find a hidden pattern anywhere worth finding, it is inside the human body.
It is an audacious sentence. It is also the entire company, compressed. Everything else - the founders, the funding, the cancer tests - is downstream of that one promise.
Here is the uncomfortable secret of the big-data era: collecting data turned out to be the easy part. Genomes, clinical records, spectra, sensor streams - we drown in the stuff. The hard part, the part nobody solved, is finding the meaningful relationships hiding inside it. As datasets pile up extra dimensions, the number of possible combinations explodes past what brute force can chew through, even when you point a supercomputer at the problem.
Conventional tools cope by narrowing the question. You decide in advance what you are looking for, then go look. Efficient, tidy, and quietly self-defeating - because the most valuable discoveries are precisely the ones you did not know to ask about. The interesting pattern is the one outside your hypothesis.
Pattern Computer's bet is that this gap - between the data we have and the patterns we can actually find - is the defining bottleneck of modern science. Close it, and diagnostics, drug discovery, and genomics all move at once.
The company is chaired by Mark Anderson, a technology and economics forecaster who built a reputation - and a newsletter, the Strategic News Service - on a slightly unnerving habit of being right. He publicly called the 2007-2008 financial collapse and the 2014-2015 oil price crash. His method, by his own account, was pattern discovery applied to the world. Pattern Computer is, in effect, that instinct turned into a machine.
He did not build it alone. Among the co-founders is Michael Riddle, who co-founded Autodesk and wrote the original code that became AutoCAD - a man who has watched one revolution in computing happen up close and apparently fancied another. Brad Holtz carries the wonderfully unbureaucratic title of Chief Nexus Officer; Ty Carlson rounds out the technical leadership as CTO.
The founding bench: one man who predicts crashes, one who helped invent how the world draws, and two who keep the engine running. Suspiciously few for a company promising a revolution.
Their wager is contrarian in the most literal sense. While the industry pours billions into ever-larger models that recognize ever-more, Pattern Computer chose the harder, quieter problem - discovery - and decided to prove it on the highest-stakes data there is.
The flagship is not a chatbot or a dashboard. It is the Pattern Discovery Engine - a proprietary blend of mathematics, software, and in some cases custom hardware, built to comb high-dimensional data for higher-order relationships and hand scientists something rare in modern AI: explainable, testable hypotheses. Not a black box that says "trust me," but a collaborator that says "look here, and here's why."
That philosophy shows up in the company's work. Pointed at Namida Lab's biomarker data, the engine surfaced previously undetectable combinations of signals and improved the accuracy of a breast cancer screening test built from, of all things, tears. Pointed at the problem of pandemic testing, it paired Raman-style spectroscopy with machine learning to read infection straight from a saliva sample - no reagents, results in about fifteen seconds.
Caption: yes, the cancer screen really does run on tears. Biology has a sense of humor; Pattern Computer just learned to read the punchline.
What can you actually do with all this? If you are a diagnostics company, a genomics lab, or a drug-discovery team sitting on data you suspect is hiding something, the engine is built to find what your own tools cannot - and to explain it well enough that you can act on it.
Skepticism is the correct posture toward any company that says it can do what supercomputers cannot. So here is what is on the record, rather than on the pitch deck.
Bars are illustrative of reported result times, not a controlled head-to-head. The point survives the asterisk: this is a different order of magnitude.
The Namida Lab partnership is the cleanest proof point: a real diagnostic test, made measurably more accurate by feeding it through the engine, with a stated plan to extend the approach to prostate, colon, ovarian, pancreatic, and other cancers. It is one customer, not a hundred - but it is the right kind of one.
A general-purpose discovery engine could chase money in finance, logistics, or advertising. Pattern Computer chose medicine first, calling it the most daunting yet most immediate place to do something that genuinely changes lives. It is a revealing choice. The hardest data, the strictest scrutiny, the slowest sales cycle - and the one arena where a hidden pattern can mean a diagnosis caught early enough to matter.
The next decade of science will not be limited by how much data we can gather - that problem is solved, embarrassingly so. It will be limited by how much of it we can actually understand. Every field now generates more measurements than its experts can interrogate. The lab that can reliably surface the patterns no one thought to look for does not just win a market; it changes the pace of discovery itself.
Pattern Computer is small, deliberately quiet, and unproven at scale - a fair charge. But the wager is precise, the early evidence is real, and the team has done improbable things before. Whether they become the engine room of a new kind of science or a fascinating footnote depends on the next handful of partnerships. Either way, the question they are asking is the right one.
So picture Friday Harbor again. The ferry, the orcas, the unremarkable office at the end of the map. Inside, a couple dozen people are betting that the most important discoveries are the ones still hiding in plain sight - in the spectra, the genomes, the tears - waiting for something patient enough to notice. Pattern Computer is trying to build the thing that notices. That is the whole company. It might also be the whole point.