Perhaps I'm in the minority in that I DO think the code is "crackable" -- but maybe not by us humans or at least not by those unwilling to get eyeball-deep in some very sophisticated math and computer science. There are so many patterns that I notice in my son's blood sugar measurements, but the issue is that they happen over varying scales (hours, days, weeks, months, etc.) -- and recognizing which pattern you're dealing with, is, in my mind, a statistically rather difficult challenge for us humans. We tend to overgeneralize and be too quick to see patterns where one does not exist, and we have a strong bias towards the most recent events -- which may be helpful for diabetes in some ways, but perhaps counterproductive in others. On the other hand, AI can beat the world's best Go Players and can deal with some pretty hairy crazy stuff, so I have a hard time believing it can't deal with BGs, whose patterns are to a rough degree governed by some fairly simple linear ordinary differential equations.
However, I actually am quite optimistic that artificial intelligence can recognize patterns in blood sugar and that within the next five years, someone will have at least a working prototype of a deep-neural-network-based artificial pancreas. Until insulin gets faster, that way lies the future, IMO.
I'm probably like you in that it KILLS me to not understand the patterns. This is not just about BG management, but a kind of ticklish curiosity that keeps me up at night. I try to let it go as spending too much time on it is a recipe for burnout, but it just is tough because I am by nature a super analytical person and spent a lot of my time creating mathematical models of biological systems in the past. So it's just hard to let it go. But I have to cause my husband doesn't find it nearly as scintillating to discuss BG trends and theories.
One thing I've noticed is that , even though my son is on an artificial pancreas that is essentially responding to blood sugar trends in the moment, before that I was relying on my pattern recognition and gut instinct and both were equally effective in terms of lowering his a1C. So it emphasized textis_emphasized text_ possible to recognize patterns and make the most of them. But the difference is in the quality of life, which is huge. Me constantly thinking about my son's BGs was a recipe for mental burnout and exhaustion, whereas the AP is achieving comparable overall results, but with a lower TDD, less averted hypos and a much higher quality of life for me. So in other words, maybe his A1C would have been a 6.8 without the AP, but if you look at the graphs they look totally different.
Our next goal in leveling up is to somehow combine pattern matching and data analysis with the APs algorithm in constructive ways. Not sure it's possible as they're sort of different approaches and sometimes cancel each other out, but it certainly would be worth it if we could achieve it.