The treadmill said 15.1 km/h. My Garmin said 12.3 km/h. Apparently, we’d been doing two completely different runs on the same machine.
I was there. I ran it. My legs remember it rather well.
And somehow that disagreement sent me down a rabbit hole that lasted a year and a half. Watches, apps, sensors, spreadsheets. All because I wanted the numbers from a run I’d actually completed.
From one watch to another… and another
This nonsense started in April 2025. Yes, really.
First it was Life Fitness treadmills, a Mi Band 9, Apple Health and MapMyRun. The chest strap showed my heart rate beautifully on the treadmill screen. Export the workout, though, and apparently my heart had taken the day off.
The treadmill knew how fast I was going. The chest strap knew what my heart was doing. Getting both into the same workout, however, seemed to require a diplomatic summit between devices that refused to speak to each other.
Then came the older, stubbornly offline treadmills at Fitness First. We tried iSmoothRun, more recording combinations and eventually an Apple Watch Series 11. Roughly S$1,200 later, I returned the watch.
Lovely watch, actually. But after spending that much money, I thought asking it to let me correct an indoor run wasn’t particularly unreasonable.
So we started looking at things to attach to shoes. Stryd, COROS POD 2, even NPE RUNN, which attaches to the treadmill. Then came the smart-shoe archaeology: Nike+, Adidas miCoach and assorted attempts to turn footwear into an information system.
Think about that for a moment. We were considering buying a computer for my shoe because the treadmill and the watch couldn’t agree on how far I’d run. Sisyphus, but with Bluetooth.
18 May. Two records. And a watch with opinions.
The two runs
Treadmill · Expected: 12:55 (4:16/km)
Treadmill · Expected: 4:00 (15 km/h)
On 18 May 2026, I first ran 3.03 km in 12:55, averaging about 4:16/km. Then, after a break, I ran 1 km in 4:00. That’s 15 km/h. Two personal records, better than the ones I’d set at eighteen. More than two decades later! I don’t think I need to explain how that felt.
Garmin had other ideas. Both original recordings were wrong. For the 3.03 km run, it reported 4:48/km instead of 4:16/km. Thirty-two seconds added to my pace for every kilometre. For the later 1 km, the original FIT lap recorded 907.62 m in 4:07, or about 4:33/km. The treadmill had shown 1 km in 4:00. Two personal bests, neither recorded properly. Both activities have since been corrected in Garmin Connect.
Garmin was still where I wanted my running history to live. It had the laps, pace, heart rate and everything else. I just wanted to correct the activity there. Not create another spreadsheet that politely disagreed with the original record forever.
So I gave ChatGPT the keys
So we built Garmin by geekist.co. One connection between Garmin Connect and ChatGPT, or another compatible AI assistant. Sign in to Garmin, complete verification if needed, and suddenly we can actually ask questions about the runs instead of playing detective with screenshots.
Reading the data was the first bit of fun. We could inspect laps, heart-rate samples, personal records, effort, feeling and activity photos where Garmin had them. We could look at sleep, HRV, stress, steps, weight and Body Battery too. And instead of trying to remember what happened last Tuesday, we could ask for trends across weeks of running. Distance, pace, frequency, duration. The lot.
Then came the part I really wanted: writing back. Correct supported activity details, add a manual run, import FIT, GPX or TCX files, or build a structured workout with the assistant. We can preview the workout before saving it. Changes need separate permission and confirmation, and deleting something needs explicit approval. I have spent far too much effort recording these runs to let an overeager assistant casually erase one.
It’s free. We can also compare Garmin with other sources, including Google Sheets, when a run needs a second opinion.
The treadmill photograph is still my favourite example. Take a picture of the console and paste it into ChatGPT. The assistant can read the distance and time, compare the figures with Garmin, and propose a supported correction. I review the change before it’s saved. But what started as an argument with a treadmill has turned into something much bigger. We’ve put the features, updates and feedback channel at garmin.geekist.co.

Come run with me ↗
If you want to follow my runs and occasional ridiculous experiments, come run with me on Garmin Connect. I’d love to hear about the runs that sent other people down similarly ridiculous rabbit holes.
Leave a comment with the watch and treadmill you use, what went missing and what you’d like to fix. Those stories are far more useful to us than guessing which feature to build next.
Personally, I’d rather be running. 🙂 Garmin Your Garmin data, in conversation Your watch has recorded the details. Connect Garmin to your AI assistant and start asking better questions about them. What you can do 01 Explore activities, laps and daily health 02 Compare running trends and personal records 03 Preview and approve running or cycling workouts 04 Correct supported activities and import files It’s free. Changes require your approval. Independently developed by Geekist. Not affiliated with or endorsed by Garmin.
Get more from every run.
