So, I have included a mildly predictive low glucose suspend that forecasts not on the AR equation, but by averaging velocity for the last 4 readings. We are just running along side, not actually using, but it is performing reasonably well. I have only been running it for a couple days, but so far no false positives. Even if it does have a false positive, basal resume once the predictive threshold is no longer met. Its best catch so far is 20 minutes early, but we only had a few very mild lows.