Non-invasive glucose monitors will take a while
theverge.com
theverge.com
I'm using the Freestyle Libre 3. It's the size of a penny on my upper arm. I don't have to replace it for 14 days, so I'm only really aware of it 2x per month.
It transmits the signal to my phone via bluetooth. So essentially it's like your phone reads your blood glucose live but every 14 days you need to wipe your upper arm with an alcohol wipe and stick a new one in. The process is completely painless and takes 1 minute or so once you've done it a few times.
Any doctor can prescribe you a CGM and they're like $70-100 per 14 day period, depending on brand/location/insurance. If you're diabetic your insurance will likely pay for it.
Should you be on the fence I absolutely recommend you just get one. Pay it out of pocket just to learn how you respond to the things you eat and drink.
But there's more to be done. They don't measure blood insulin, but the interstitial fluid, which responds more slowly. And a new monitor takes an hour to establish its baseline, which is an hour that your closed loop isn't complete. (I gather that the newest Dexcoms take just 30 minutes.)
A version that used just a smartwatch would be amazing, especially if it really was measuring the actual blood level. But making that sure enough to be part of a closed loop is a massive, massive hurdle. It's controlling the delivery of a medicine where both overdose and underdose are dangerous, and that means extremely high levels of precision and proof.
If this was on by default in every Apple watch we'd have 20 million fewer T2Ds the next week.
The problem is that the FDA is basically forbidding anything that could even remotely be "accidentally abused" by T1Ds. To even a consumer-focused fitness CGM/watch function would have to comply by the most stringent FDA regulations like medical devices marketed specifically to T1Ds, unless you somehow modify it in a way that would prevent T1Ds from "accidentally using it." I've heard you'd need to e.g. delay the feedback by 24h, which makes it pointless.
It won’t tell you any kind of absolute measurement, maybe just some sort of rough guide to how fast it’s going up/down of if things are relatively stable.
Enough to prove useful and know when you need to test with a real test, but not a full testing solution itself.