I was surprised when my new endocrinologist mentioned this idea. As a software developer, I would not trust a device like this - the risk is too high - although automating reporting and notification would be plausible for me. That said, I ran into another Type 1 at my gym - I noticed her by the Omnipod sensor on her arm - I was intrigued. I have used a Dexcom CGM, and before that used a Medtronic insulin pump for its CGM, and have often told myself that I would not use a pump until there were no more tubes. I guess that day is upon us.
I understand people holding this opinion. Loop, OpenAPS, and AndroidAPS are all experimental systems, not FDA approved, and anyone who builds one assumes all liability in their use. I have been using Loop for 21 months now. I feel that my risks were higher prior to my adoption of Loop. It does a great job of taking me safely through the night and dependably delivers wakeup blood glucose of 70-99 mg/dL.
Since I starting using Loop, my time in range has significantly improved, time hypo reduced, and my overall BG average lowered to below 100. The open-source nature of these systems means many different people can examine the code for any flaws. I am impressed that other software developers have examine this code and have then placed their children on these systems. Meanwhile, millions of hours of use have been logged and the code is continuously being modified. Safety seems to be their highest priority.
Your skepticism is healthy and commercial FDA-approved systems will be available in the next few years.
Unhappily, in order to gain FDA approval, it seems that unrealistic goals and methods have to be built in… vis a vis the latest Medtronics 670g. Everything has to be safe for the lowest common denominator.