What your Nightscout data has been trying to tell you.
Paste your site. LoopedTune reads the last few weeks and comes back with the two or three trends that matter most — nights with lows, corrections that overshoot, meals that never come down, days that behave differently — each one with the actual days behind it, so you can see it rather than take it on trust.
How it decides what to say
Every note has to survive a numeric check before you see it. Fixed rules catch the known patterns; a model then scans the whole period for anything else and proposes candidates — but each candidate is recomputed from your data and dropped if it doesn't hold. Overnight drift only counts nights with no food or bolus after 20:00 and needs at least five of them agreeing. Correction strength only counts boluses with no carbs three hours either side. Meal excursions need four or more logged meals in the same slot and look at four hours, not two, so timing problems don't masquerade as ratio problems. Lows always come first. If nothing crosses a threshold, it says so rather than inventing something.
Notes name a direction, never a dose. The point is to tell you where to look and what to bring to your care team, not to hand you a profile to paste into a pump.