More from this creator
Other episodes by Kitty Cat.
More like this
If you liked this, try these.
Transcript
The full episode, in writing.
A continuous glucose monitor is not actually measuring blood. It is measuring interstitial fluid — the watery layer that sits between cells, just under the skin. A coated electrode pokes through the dermis, an enzyme reacts with glucose molecules in that fluid, and the reaction generates a tiny electric current proportional to the glucose concentration. A transmitter on the surface of the skin reads that current every one to fifteen minutes and sends the value to a phone or a receiver. Because glucose has to diffuse from capillary blood into interstitial fluid before the sensor sees it, the reading lags real blood glucose by five to twenty minutes — longer if the wearer is moving fast, eating, or exercising.
The first CGM was approved by the FDA in 1999. It was a Medtronic MiniMed device that recorded a reading every ten seconds and reported a five-minute average, with a sensor that lasted 72 hours. Dexcom's first system arrived in 2006 — also 72-hour wear, with a five-foot wireless range. Abbott's FreeStyle Navigator followed in 2008 with a five-day sensor. Each generation extended sensor life and shrank the receiver. By 2012, Dexcom had pushed wear time to seven days and transmission distance to twenty feet. By 2015 the data went straight to an iPhone.
The pivotal moment was September 2017. The FDA approved the FreeStyle Libre, the first CGM that did not require a fingerstick blood-glucose sample to calibrate. The original Libre was a "flash" monitor — the sensor stored eight hours of data and the user waved a reader near their arm to download it over near-field communication. No alarms, no live stream. The FreeStyle Libre 2 added Bluetooth, fourteen-day wear, low and high glucose alerts, and a one-hour startup time. The Libre 3, released in 2022, transmits every minute. In June 2018, the FDA approved Senseonics' Eversense, the first fully implantable CGM, with a sensor inserted under the skin and a removable transmitter worn on top. The original Eversense lasted 90 days; the Eversense XL went 180; in October 2024 Senseonics launched the Eversense 365 in the United States, the first one-year CGM.
Two companies dominate. Dexcom and Abbott account for essentially the entire global market — Dexcom alone reported more than three billion dollars in revenue in 2023. China-based Medtrum and a long list of noninvasive challengers — Movano, Know Labs, HAGAR, Profusa, Glucomodicum, Liom, Occuity, DiaMonTech — have prototypes, patents, partial regulatory approval in places like Saudi Arabia or Europe, and almost no commercial traction. Apple has been chasing a noninvasive sensor in the Apple Watch since at least 2023; as of early 2026 it remains at the proof-of-concept stage. Samsung has announced spectroscopy-based glucose monitoring repeatedly without shipping it. The hard physics problem is that glucose absorbs strongly in the mid-infrared, where wavelengths are measured in micrometers — radio frequency and millimeter-wave approaches keep failing to detect it through the skin.
The non-diabetic consumer market opened when both market leaders launched over-the-counter versions. Dexcom released Stelo, a two-week disposable sensor sold without a prescription, aimed at people with type 2 diabetes not on insulin and at the broader wellness audience. Abbott released Lingo for the same demographic. Subscription startups like Levels and Nutrisense had already been brokering prescription Libres to non-diabetic biohackers for years; the OTC clearances cut out the prescription middleman. The pitch is that watching your own postprandial glucose curves teaches you which foods spike you and which don't.
The science behind that pitch is thin. A 2012 Cochrane review found limited and conflicting evidence that CGMs improve outcomes even in poorly controlled diabetes. A 2024 meta-analysis of 25 randomized trials found that CGM-based feedback produces only modest reductions in A1c and modest gains in time-in-range, in both diabetic and non-diabetic users. The sensors themselves miss low blood sugar — a 2021 systematic review specifically flagged that CGMs are not accurate enough to detect hypoglycemia, which is why NICE in the UK still recommends confirming any low reading with a fingerstick. Sleeping or sitting on the sensor causes "compression lows" — false hypoglycemic alarms triggered by physical pressure on the electrode. In February 2025, the FDA issued a public alert that smartphone-paired CGMs were silently failing to deliver critical low-glucose alerts because of phone notification settings, and people were missing them.
The most consequential recent development is on the data side. In January 2026, a team led by Guy Lutsker published a paper in Nature describing a foundation model for CGM data — a transformer architecture trained on ten million glucose samples that could predict the future onset of diabetes and cardiovascular conditions years in advance from the shape of the glucose curve alone. Closed-loop diabetes systems already exist: an insulin pump reads the CGM, an algorithm calculates a dose, the pump delivers insulin without the user touching anything. The Medtronic MiniMed and Tandem Control-IQ are the commercial versions. The open-source community runs OpenAPS, an unregulated artificial pancreas that diabetics have built and shared since 2014. The hardware that started as a 72-hour sensor in 1999 is now the input layer to a fully automated organ.