670g unbiased detailed review by Gary Scheiner, author of "Think Like a Pancreas" book

Here is the good, the bad and the ugly of the 670g according to Gary Scheiner, author of “Think Like a Pancreas” book. Your thoughts please. There is a good description of the algorithm used by this system. Is this system using a similar algorithm used by the Loop system that you are using @Terry4?

I like what I see with this system despite it’s weaknesses. If you can tolerate a good consistent control, I think this will work for you. Take note that it’s not being aggressive by design to satisfy FDA safety concerns. Health-care politics.


There are some people who don’t consider a 120 BG target as “tight control”…


True. 120 target is not tight. I reworded the OP. From TIGHT CONTROL TO GOOD CONSISTENT CONTROL . I had in mind flat lining as opposed to camel-humpback trending. I remember reading that hemoglobin glycating happens during high blood glucose spikes as well as out of target duration. Hence the importance of the measure of time on target. But I do invoke the YDMV. LOL

I prefer a system where the user can set the upper and lower limit, instead of a fixed target of 120.
But having a pump that can adjust your basal automatically is still a step forward.
In a few years, we should see voice command with artificial intelligent pump.

"Hey AI PUMP, I just ate 30 grams carbs with 50 grams roast chicken. Shoot for 70-120 BS range. OK? Give me a BS report in 2 hours.

AI Pump, tell me my projected A1c?


No, not all. Loop allows customization. I set my BG target to 83 mg/dL. I can set a temporary target a little higher for exercise or a little lower when I want to deliver more insulin. I use it like the gas pedal in a car.

Loop allows me to set various insulin sensitivities and segment the day with them. Loop doesn’t force me to distort the actual meaning of insulin duration in order to allow the user to change prospective performance like the 670.

Since I’m not algorithm literate, I can’t compare the two algorithms but Loop is unimpaired by regulatory niceties. It doesn’t send me into a redundant button-pressing circus every time I just want to add some insulin. Loop designers built in many safety provisions including falling back to the settings programmed into the insulin pump when the system fails. The most durable command Loop gives is a 30-minute temp basal rate.

Loop’s open source nature and citizen programmers, free of regulatory burden, can continuously iterate and as a result evolve more quickly than any commercial offering. I feel like I am enjoying a system that’s ten years in the future for the commercial systems.

I see the 670G system as a baby step toward an automated system, but it’s a crucial baby step. The commercial companies will deliver this technology to a much broader demographic than any of the DIY automated insulin delivery systems can.

I wish that the Medtronics of the world would open up their code and allow the talent of the open source do it yourself movement to help all of us more quickly.

The JDRF supports this as do some of the commercial diabetes tech players.

1 Like

My average bg is less than that, and 120 is also not my target, either. :slight_smile: It’s 100. And often I’m in double digits; not 3. (thanks to Dexcom)

Love your Loop system but not easy to build. Would be nice if I can go to Best Buy and tell the geek to build me one. LOL. With Loop’s flexibility, FDA will probably not approve a system this sophisticated. (Diabetic euthanasia possibility).