The central laboratory has been the unquestioned architecture of molecular diagnostics for three decades, and almost every assumption built into it is a liability. Samples must be collected, boxed, transported, refrigerated, mixed, loaded, and run on equipment the size of a desk, and interpreted by a trained technician. Each of those steps is a cost line, and each is a point where a result can quietly go wrong without anyone knowing. Dr. Leonard E. Weisman, who leads the development of the LabReady Multi-Point quantitative polymerase chain reaction (qPCR) AI device, has spent his career watching that chain fail in ways the final report never discloses. His argument is blunt: the laboratory workflow was never the science; it was the packaging, and the packaging can be stripped down to something that fits in a palm.
Subtraction As Engineering Strategy
What Weisman describes is less an invention than a sequence of deliberate removals. Refrigeration goes, replaced by freeze drying. Chemical mixing goes, because the reagents are preloaded in the cartridge. The separate collection device goes, absorbed into the unit itself. The large thermocycler built to run 96 or 384 wells goes too, and with it the engineering burden of heating and cooling at scale. The wells are small individual reaction chambers or cavities in a plate. “We’re generally looking for one set of targets, and so we’ve miniaturized that to one well, which makes the heating and cooling much easier,” he says. The result is a six-inch disposable device that fits in a pocket or a backpack.
The logic here cuts against how diagnostics companies usually think about capability. Throughput has been the industry’s proxy for sophistication, and the big analyzers earn their keep by batching hundreds of samples at once. But batching only makes economic sense when samples travel to a fixed point, which is precisely the constraint Weisman is attacking. Strip the requirement to centralize, and running 384 wells becomes irrelevant overhead. A single-well device answering a single question, on site, in minutes, is not a degraded version of the lab. It is a different machine solving a different problem, and the engineering gets simpler as it gets closer to the patient.
The Failure Nobody Records
The most consequential part of Weisman’s case concerns a failure mode that leaves no trace in the data. Samples collected in the field sit in boxes exposed to heat or cold, waiting for a pickup that is supposed to arrive within hours and sometimes does not. By the time the specimen reaches the bench, the conditions of its journey are invisible. “There’s no way they can identify whether a sample that arrives has been handled appropriately,” he says. “And so they just test. If the sample’s there, they tested.” The result is issued with full institutional authority, and the degradation that occurred in transit is simply absorbed into it.
That is the quiet scandal of distributed sample collection, and it is worth sitting with what it implies for accountability. A laboratory can be flawlessly run and still produce a misleading answer, because the error entered the system before the sample arrived and no step in the workflow is designed to catch it. Weisman frames elimination of transport as the direct fix: a disposable unit designed to return a result on site in minutes shortens the window in which a sample can degrade. The downstream consequence matters more than the logistics. In principle, a clinician with a verified answer at the point of care has less reason to guess at therapy while waiting on a distant lab, and less reason to act on a number whose provenance nobody can vouch for.
Putting Quality Control On The Board
Moving PCR out of the lab raises an obvious objection: the technician did more than run the machine, because the technician was also the quality control. Someone with training watched the numbers come off the instrument, confirmed the controls behaved as expected, and judged whether inhibition or spurious signal had compromised the run. Weisman’s answer is to move that judgment onto the device itself. The AI runs on edge, on the board, managing the process, and verifying the result. “It tells you whether the control is positive or negative, as it should be. Is the information that you’re receiving from the device consistent with no inhibition and no false information? Basically, it’s a quality control system.” The device is designed to handle in seconds what a technician once did by inspection.
This is the step that makes decentralization credible rather than reckless, and it reframes what onboard intelligence is really for. The AI is not interpreting disease or recommending treatment. It is doing the unglamorous verification work that made laboratory results trustworthy in the first place, and doing it at the point of collection instead of at the end of a chain of custody. That shift is intended to lower the skill floor far enough for the device to work on a battlefield, in a restaurant, in an office, or in a farm shed. Weisman is explicit that workflow, not specification, decides which devices survive contact with the field: “Collect the sample, put it in a device, push a button, and get your results in 15 minutes on your smartphone. That’s really what we’re looking for.” Plenty of point-of-care products have been announced with impressive sensitivity figures and have never left the brochure, because they still needed a cold chain, a trained operator, or external support. Simplicity is the competitive moat. The near-term applications Weisman names, including healthcare, farming and agriculture, and cancer care, are narrower than the opportunity he sees, which extends to traumatic brain injury and, in principle, to anything detectable through DNA or RNA. Deployability matters more than laboratory-grade claims, because the technical problem was largely solved years ago. What has not been solved, until the workflow collapses to a single button, is getting the answer to the place where the decision gets made.
Dr. Leonard E. Weisman shares additional perspective on molecular diagnostics, point-of-care testing, and edge AI in clinical decision-making on his LinkedIn profile.
