Quick verdict, then I’ll earn it: nobody has run a trial showing a continuous glucose monitor (CGM) makes a healthy athlete perform better. Not one. As a one-month lesson in how your body handles sugar, it’s worth it. As a gel timer for a marathon, it’s potentially theater.
I wore it for 15 days, and it changed exactly nothing... yet!
What I put on my arm. A Dexcom G7 15 Day sensor, back of my left arm, September 17 to October 2. Setup took a minute — alcohol wipe, press the applicator to the arm, push the button, done. It connected continuously and pushed data to my iPhone and Apple Watch Ultra 2 in real time for the full two weeks without a dropout, and once I connected the app to Apple Health, every workout fed in automatically, stamped with the reading at the moment it started. I don’t have diabetes, which is why the first thing I had to learn was how to turn off the high and low alarms. Very useful for the people this device was built for. Not for me.
How I got it: I asked my doctor - who is also an athlete, a triathlete - about the best way to fuel long runs without running myself into glucose depletion, and without taking carbohydrates in the “recommended amount” from a company that doesn’t know me. We agreed a sensor was a reasonable way to see what my body does with a carb gel. That conversation led to the script.
What I do have is a sweet tooth, a rehabbed Achilles, and a first NYC Marathon in 2027 that I’m working back toward one long run at a time. So the question I put on my arm was simple. When I start taking Maurten gels on long runs, can this thing tell me when to take the next one?
What the line did. The numbers I’ll print: my average across all 15 days was 108 mg/dL. Most mornings before breakfast it sat in the low 90s. Ninety-nine percent of the two weeks was between 70 and 200 - the range I set in the app - and it never once dropped below 70, awake, asleep, working out, or running. For scale, the biggest study of healthy people my age - 560 of them, average age 59 - averaged about 115 across the day and spent roughly one hour in seven above 140. Nobody’s line is flat. Mine ran a little lower than theirs, and I eat candy. I’ll take it. And the fittest people ever put on these sensors run high, not flat: when researchers sensored an elite European football squad, the players on the pitch averaged 159 mg/dL during the match against 133 for the reserves on the bench - the same meal, the same morning, the only difference was ninety minutes of competition. Stress and effort push glucose up in perfectly healthy bodies. A number over 140 is not, by itself, a verdict on you. I checked it against a fingerstick exactly once; the two numbers were a couple of points apart. Once is not an accuracy study. It’s one honest sentence.
One caveat from Dexcom’s own pivotal data that nobody tells a first-time wearer: the sensor is least accurate on the first three days and the last three - exactly when you’re staring at it hardest. The researchers who do this for a living treat day one as noise, and so should you; nothing in the numbers above leans on September 17. It took me a day or so to learn to read the app. After that, every day looked remarkably like the one before it. Every day but one: the night I had more than usual to drink, the line looked different, my sleep looked different, and so did the next morning. My somewhat educated hypothesis is that the alcohol altered the sleep and the sleep altered the glucose. I won’t pretend I know more than that.
The candy does exactly what you thought. I love gummies, fireballs, Twizzlers. I ate them. The line went up. What I didn’t expect was how fast it came back down - 30 to 45 minutes most times, depending on what and how much. Which leads to the one thing from the whole two weeks that genuinely stood out: eat and then work out, and the line comes down faster and smoother than eat and sit. Not a little faster. Visibly. The sensor’s best trick wasn’t telling me what food does. It was showing me, in my own trace, what movement does to food.
Someone close to me showed me the real effect first. She wears one because her doctor asked her to. Mine was curiosity and a script. That asymmetry matters, so I’ll say it plainly: hers is medical, mine was elective, and nothing I learned about gels applies to managing a condition. Her numbers are hers, and they stay off this page.
After weeks of looking at her own line, she found that a walk after dinner and dessert managed her glucose far more evenly than sitting. Her realization, in a sentence: exercise helps her body handle glucose better. I wasn’t wearing a sensor the evening we walked together - the lesson came from what her readings had done on the nights before versus that one. Weeks later, I watched the same shape show up in my own trace, more than once, every time I moved after eating - a walk, a workout, didn’t seem to matter.
That’s not a hack; it’s old physiology. Working muscle pulls sugar in without needing insulin. A 2025 crossover study put a number on the timing: a 10-minute walk started right after a glucose load cut the peak from 182 to 164 mg/dL, while a longer walk that started 30 minutes later didn’t touch the peak at all. Twelve people, pure glucose, not a real dinner — but it’s the thing she saw on her own, without a lab. The sensor just made it visible in your own body, which turns out to be different from reading it.
The long run. Six runs while wearing it, three to six miles each — I’m still building back, so nothing long enough to need fuel. I took one Maurten Gel 100 on the six-miler, mostly to see what the line would do. A slight rise, very similar to candy but smoother - no cliff - and it corrected quickly. Never below 70 on any run. Nothing pressed on the sensor (the only thing that ever caught it was a T-shirt coming off). I never looked at the watch mid-run — I felt it would have been distracting, and the phone app let me scroll the previous 24 hours almost minute by minute afterward. I did not overlay the trace on Strava to compare my HRV and other stats. That’s next.
Here’s why “next” matters. The sensor doesn’t read blood. It reads the fluid between your cells, which lags blood by five to fifteen minutes — longer when things are moving fast. Do the math on a gel: it takes 15 to 30 minutes to get into the blood after ingestion (slower while running hard), then another 5 to 15 minutes to reach the sensor. The G7 updates every five minutes, and the copy that lands in Apple Health is - Dexcom’s own words - “on a 3-hour delay, not in real time.” By the time the watch shows a dip, it’s reporting on a decision you needed to make 20 to 40 minutes ago. This is why every evidence-based fueling plan is scheduled and prophylactic: the physiology is slower than the race.
In August I wrote that “you never needed the sensor, you needed the staff” — the person paid to read the number and change what you do tomorrow. For glucose, the staff already did the reading. The nutrition consensus handed everybody a schedule: 30 to 60 grams of carbohydrate an hour for efforts under two and a half hours, about 90 an hour beyond that, glucose and fructose mixed, with your gut trained to handle it. A Maurten Gel 100 is 25 grams — call it two and a half gels an hour at the low end, roughly $38 of gel across a four-hour first marathon. I haven’t changed my cadence because I haven’t had a run long enough to have one. My gut is still the experiment.
So did any of it change a decision? Straight answer: not yet. I changed nothing on purpose - same food, same training - because I wanted a baseline before I started pulling levers. The next sensor is where I test changes: Strava and HRV overlays on every run, real fueling changes, longer runs. And because I’d rather do this properly than as a guy with a sensor and a Substack, I’ve written to three researchers who work with elite endurance athletes on fueling to ask whether one of them would run this as an actual study. I’ll report back either way.
That’s the worth-it half. Now the half I actually care about.
The sensor is watching the wrong tank. Here’s the mechanism. When you run long, your blood glucose is only about 20 to 30 percent of the carbohydrate you burn. The rest comes from glycogen stored in the muscle - and that’s the tank that empties at mile 20. The CGM can’t see it. The most rigorous independent review of this, published last November, says it flatly: there is “insufficient evidence to recommend the use of CGM to determine fuelling strategies,” and the causal link between glucose (short of actual hypoglycemia) and performance “remains absent.” Those aren’t skeptics from the cheap seats. They’re the sports-nutrition people.
And it gets worse right when the race gets hard. In steady, easy running, the sensor is decent. But in one study of well-trained men, once lactate climbed past 4 mmol/L, blood glucose rose, and the sensor didn’t track that accurately. Dexcom itself says something pressing on the sensor can throw the reading — a hydration vest strap on a hot day, in polite terms. And in the largest athlete dataset that exists — twelve thousand people, built by the company that later died — about one minute in eight during exercise read “low.” Those people were fine. A sensor low mid-run is not the bonk.
Somebody actually tried “eat when the sensor says so.” One trial, 2025: twelve recreationally active men, 75 minutes of easy cycling, fed on a schedule or whenever the CGM told them to. The CGM group took 18 grams of carbs an hour versus 51. Heart rate, effort, lactate — identical. Nobody measured performance. Four of the twelve sensors actually fell off in the study. That is the entire body of evidence for the feature I wanted. And 18 grams an hour in a marathon isn’t a strategy; it’s how you meet the wall on purpose.
Ahead of the wave. The company built for exactly my question is dead. Supersapiens - the athlete-CGM startup Ineos and Jumbo-Visma used in training - shut down in early 2024: no FDA clearance, one sensor supplier, roughly €1.3 million in revenue, and data that, per DC Rainmaker’s post-mortem, wasn’t reliable enough to make training decisions on. He’d worn it for three years and found it “only functioned reliably for steady-state efforts exceeding 90 minutes.”
I’ll put on the investor hat for a second and follow the 2026 money. Dexcom bought a nutrition-coaching company, rebuilt its over-the-counter sensor into an AI food logger, and wrote a check into a weight-management app alongside Google’s venture arm. None of those moves have a runner in them. The number a runner actually wants is lactate, and as of this month you cannot buy a continuous lactate sensor you’d trust — one is on preorder with the company grading its own homework, the rest are waitlists, and Dexcom’s roadmap through 2028 doesn’t have it. The week I was wearing mine, the FDA cleared a needle-free glucose patch for the pharmacy shelf. It doesn’t even give you a number — just a color. That’s the wave: fewer needles, more analytes, aimed at prevention and the GLP-1 patient. It is not aimed at either me or the people who study this quietly moving the goalposts the other way: the emerging view in the exercise-glucose literature is that for a healthy, active body a little higher may be better than the flat 70-to-140 line the wellness apps sell — which makes a sensor sold on “spikes” a strange product to hand a runner.
Which is the asterisk on the title. I’m not ahead of a product — the product is turning away from me. I’m early to the habit: reading what my own body does with sugar and with movement. The habit is the one part nobody can sell me.
The scorecard. Five days into the wear, I wrote the criteria down so I couldn’t move the goalposts:
The theater criteria, written down now so I can’t move the goalposts: if by day 30 the trace hasn’t changed a single decision - or if all it told me is that candy spikes and walking helps, which I already knew - it’s Theater. If it changes my breakfast or my gel cadence and the back half of a long run feels better twice in a row, it earned its money. Once.
Grading myself in public, at 15 days instead of 30:
Changed a single decision: not yet.
Back half better twice in a row: can’t tell. I changed nothing by design, so there was nothing to feel.
Beyond “candy spikes and walking helps”: yes, one thing. How much faster fitness corrects the line — that I’ll confirm. The specifics come with the next sensor.
And the four-week plan, honestly:
Week 1, baseline, touch nothing: done.
Week 2, the dinner walk: done, more than once, and with workouts after eating.
Week 3, long runs with the overlay: not yet. Three to six miles, one gel, no overlay.
Week 4, change one variable: the sensor was off. Next sensor.
Would I wear one again? Yes, and I will. I chose the prescription G7 versus the over-the-counter Stelo for one reason: the G7 talks to my Apple Watch and connects to Apple Health, and I don’t run with a phone. That’s roughly $300 to $400 for two sensors without insurance, against $70 to $100 over the counter. If a friend asked whether to buy one, it’d be a discussion, not a yes — I’d want to know what they were hoping it would tell them. And if someone shipped a lactate sensor I could trust, I’d buy it tomorrow.
The literature can tell me what the average trained man does with a gel. It can’t tell me what I do with one at mile 14 after my breakfast. The sensor isn’t a coach. It’s a mirror.
I’ll tell you now the candy isn’t going anywhere — I just might take a walk afterward.
One question, one line. Not what a wearable showed you — what it changed. The single thing you eat, drink, or do differently because a device made you re-examine. And the honest second half: did it stick 30 days later, or did it last a week? Mine’s in the first comment. I’m collecting these, and I’ll publish the roll-up next week — no numbers, no names you don’t give me.
* Ahead of the wave means early to the habit, not the product.
This is my experience as an athlete, operator, and investor — not medical advice. The sensor came from my own doctor, a triathlete, after a conversation about fueling. I wear the sensor. I just know what it is and what it isn’t.
New to OpenTabs? Start Here.