Your pump automates insulin. You still do everything else.

Today's systems control glucose well — but only as well as the information they're given, and that still depends on the person getting a hundred small things right, every day. Dossi is designed to shift more of that work from the person to the system, keeping the familiar AID safety backbone while asking less of the human running it.

Get early access Explore features

Better inputs

Forgiving of imperfect information

Better experience

Fewer decisions left on you, all day

Better personalization

Learns your body as it changes

Dossi home screen — live glucose, prediction, insulin on board, and active extended delivery

The Problem

The math is solved. The workload isn't.

For fifteen years, automated insulin delivery has been treated as a control problem—better math, tighter loops, smarter dosing. It worked, and the systems have largely converged on glucose control. But a controller is only as good as what it's told, and it still leans on the person for all of it: the carb counts, the timing, the corrections, the settings that drift with sleep, hormones, and activity. It's why control that looks one way in a trial slips in real life—and why living with an AID system can still feel like a second full-time job.

Where the work still lands

The carb count is still on you

Every system today needs a number for every meal, and the dose is only as good as your estimate. Most meals aren't a clean number.

A meal is more than carbs

Fat and protein stretch absorption over hours, but today's systems see one carb number and one curve.

A rough guess, dosed as exact

Say “about 40 grams” and the loop treats it like a lab measurement. Being unsure isn't something the math accepts.

Your body shifts; your settings don't

Sensitivity, basal, and ratios drift with sleep, hormones, and activity—and keeping them current is usually a manual job.

Exercise is one more thing to manage

Activity can pull your glucose down for hours—but you're the one who has to remember to flag it, before it does.

Aim high to stay safe

To avoid lows, most systems hold you well above where you'd like to sit—trading time in range for a safety margin.

The thesis

A system that knows more and does more, so you do less.

Read the argument

Features

The same closed loop, carrying more of the work.

It watches, so you don't

Reads your CGM every five minutes and pulls sleep, workouts, and heart rate straight from Apple Health—no logging on your part.

It learns your body

Learns your carb ratio and dawn pattern from your own outcomes and doses on them—while keeping other context separate from learned insulin adjustments.

It thinks three hours ahead

Projects your glucose over a three-hour horizon, weighs the risk of a low more heavily than a high, and adjusts basal every five minutes—easing off when the CGM signal is noisy or stale.

It takes the meal off your plate

Log a meal by photo, a sentence, or a quick chat; Dossi estimates the carbs and macros—as a range, not a false exact—and you just adjust.

Explore features

Dashboard Demo

Nutrition

The Difference

The difference is everything around the dose

Dossi does not try to replace the AID category with something unfamiliar. It builds on the same closed-loop foundation and focuses on the work around insulin delivery: what users enter, what settings they maintain, and what decisions they have to interpret.

Capability	Modern AID pumps	Dossi
Auto-adjusts insulin from your CGM
Auto-corrects highs without a manual bolus
Adapts to your workouts automatically (Apple Health)
Models meal fat & protein, not just carbs
Log a meal by photo—AI estimates the macros
Learns carb ratio and dawn-pattern changes from outcomes
Uses health context without generic hidden factor adjustments
Shows its reasoning for every dose

This table highlights workflow differences, not proof of clinical superiority. Dossi's validation question is whether reducing that workload can match or beat standard-of-care control while improving reported diabetes burden.