Engineers at the University of Cambridge have built a chip-scale light sensor that costs around $10 and, in a proof-of-concept test, tracked the blood glucose of one volunteer without a needle. The device is a spectrometer, an instrument that reads the colors of light, and the work was published in Nature Photonics as “Optical convolutional spectrometer.” The paper reports two separate tests.
The day in tech, in one email: TLDR is the daily newsletter with links and TLDRs of the most interesting stories in startups, tech and programming.
We may earn a commission if you sign up.
The result is early. The glucose demonstration involved a single participant, the cost is the team’s own approximate figure, and the researchers say more testing is needed before anything like this could become a medical product. The broader pitch from the team, led by Chunhui Yao and Prof. Qixiang Cheng with Prof. Richard Penty among the authors, is a pocket or wrist instrument that could one day do laboratory-style measurements of the kind now confined to benchtop machines, a goal that rests on far more validation than a single volunteer can provide.
Doing the mathematics in light
Most miniature spectrometers split light with a grating or reconstruct a spectrum from many noisy measurements. According to the Cambridge Department of Engineering’s account, this one applies the convolution theorem to incoming light directly in the optical domain, using a cascade of components with repeating spectral responses, such as unbalanced Mach-Zehnder interferometers or micro-ring resonators. Tuning them shifts the response in a predictable way, and fast Fourier transforms recover the spectrum.
Chunhui Yao, the lead author, said the key insight was to go back to the mathematics, and that doing so avoids many of the limitations that have held miniaturized spectrometers back. The device sits on a silicon nitride photonic platform with onboard electronics, spans centimeters, covers near-infrared light from 1200 to 1700 nanometers and, per the release as republished by ScienceDaily, collects and processes a measurement in under a second. Richard Penty, another author, said the architecture is “simple, scalable, and manufacturable.”
What the $10 chip measured
The lab results are strong on paper. The chip classified plastics, pharmaceuticals, coffee, flour and tea with a 100% success rate, and measured concentrations in water-based and organic solutions to about 0.01%, which the team says exceeded commercial benchtop spectrometers in those experiments. It also operated reliably from minus 20 to 80 degrees Celsius, and Optics.org’s June coverage adds that sampling and local processing took under 0.4 seconds. All of these performance claims are the researchers’ own, and no independent verification is described.
The tests on the body were run under conditions designed to mimic physiology and covered four biomarkers: skin moisture, blood alcohol, blood lactate and blood glucose. The conditions were meant to reflect the body, not a full clinical setting, and the ScienceDaily version of the release reports no sample sizes or error bars for these tests and no description of how the $10 figure was calculated.
One volunteer, several days of eating
The glucose claim has two parts, both in the Nature Photonics paper. In an oral glucose tolerance test, 27 participants fasted and then drank 75 grams of glucose; spectra and fingertip blood readings were taken about every 10 minutes for up to three hours, and a model trained on the group reached a mean absolute error of 0.363 millimoles per liter.
The needle-free tracking is the second part. One participant wore the device over multiple days while eating and snacking freely, with fingertip measurements about every 10 minutes and an invasive continuous glucose monitor as an extra reference. A personalized neural network reached a mean absolute error of 0.156 millimoles per liter, and the retrieved fluctuations differed from the two references by mean absolute relative differences of 7.82% and 10.03%.
The authors state the limit plainly. Neither the group model nor the single-participant model transferred to new participants, because person-to-person variability weakened the correlation, and they say a much larger database is needed. The participants were Asian adults of both sexes across various ages, so generalizability is untested. Because the long-term model was personalized to one wearer, the result shows the sensing approach can follow glucose in one body over several days, and it leaves open whether a device could be handed to a new user and work without retraining.
From proof of concept to product
Qixiang Cheng, the project leader, said what makes the work stand out is “not just performance in the lab, but technical readiness.” His stated vision is to make spectrometry as ubiquitous as temperature or motion sensing. The work involves GlitterinTech, a start-up founded by the same Cambridge research group, and the healthcare uses described in the coverage are potential applications rather than products. Dates matter here. Cambridge says the paper appeared in Nature Photonics on April 15, 2026, and the version circulating in late October is a republication of that release, so the result is months old. The paper’s DOI is the place to check the numbers, and neither regulatory status nor commercial availability is described in any of the coverage.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
The day in tech, in one email: Get TLDR →
We may earn a commission if you sign up.
More from Morning Overview
- Lake Powell sank to a record 3,517 feet, nearing the level that stops Glen Canyon Dam’s turbines
- Seven of Earth’s nine planetary boundaries are breached, and all seven are worsening
- The NSA says three phone features should be off whenever you aren’t using them
- Hurricane Hunter radar shows four warning signs that a tropical cyclone is about to strengthen, a University of Miami study found
