Well, today didn’t have much to report. It was a meh day for me. DIdn’t sleep well, early developer meeting….just wasn’t a great day.
The developer meeting was pretty normal. Nothing huge to write home about. We are going to use ChatGPT for semantic search. Cost looks to be more than affordable but we need to make sure we have a back up as ChatGPT seems to be down 30+ hours a year currently (which is pretty high for a tech service). The other good news is, it will allow the option for people to use other languages for search and we will hopefully make a better experience for people who don’t speak English. We also discussed creating multiple languages for the website and the app. This will require us to change some code and make all of the text dynamic.
So this next part has nothing to do with Fix it Today. I had another business idea that uses AI but its in the healthcare space. I was working my nursing job and I talked to a patient with a blood sugar of almost 500. Super high. The problem is, current diabetes management (more specifically in the elderly patient population) is pretty horrible. Counting carbs and managing insulin coverage is difficult to understand. So, doctors take a more “universal” approach and tell patients to take a set amount of insulin as long as their blood sugar is over a certain amount and doesn’t count on anything like carbs.
So the idea is to use AI image recognition along with the iPhones ability to use measuring. If we can estimate the size or mass (maybe its volume?) of the foods, we can more accurately count the carbs compared to people guessing. Even if we are half way close, making it super simple for people to count carbs, we can more accurately calculate the proper amount of insulin needed to counteract the food the pt is about to eat. This could also really help in a hospital setting where we completely mismanage diabetes as well and simply just give insulin based on a pre-meal blood sugar check.
I think with future healthcare, insulin pump companies like Tandem will use an AI algorithm to individualize calculating a patients response to insulin and food intake. Imagine a patient scanning their food with an app, it calculates the carbs. The patient enters the carbs and food type (or through and API) the foods and quantity to a continuous blood glucose monitor. The glucose monitor will slowly collect data and see how the body responds to different carb types (simple vs complex carbs) and try to factor in fat. Once we have a more predictable curve of how the body will respond, we can then do the same with insulin coverage. The insulin pump can measure the response over time.
So, between understanding the carb intake, the blood sugar response based on carb intake, we can properly plan the correct amount of insulin to cover with. With time, the pump could program itself (legal issue here) and maintain more optimal blood sugars. This will drastically reduce diabetic complications and lead to a much healthier patient. Diabetes is a very hard but completely manageable disease.
Time for a provisional patent? I just want to do the app. I have no desire to create an insulin pump, glucose monitor, and pass FDA regulation.
