Why Dossi

Built to take on more of the work.

The next step in automated insulin delivery isn't only a smarter control law. It's a system that knows more about the person and the moment, carries more of the reasoning, and asks less of you every day.

The problem

Carb counting, meal timing, changing settings, exercise decisions, and second-guessing the pump add up to a constant cognitive burden. The system may automate insulin while leaving the person to operate everything around it.

The thesis

Dossi gives the controller richer context and a more forgiving interaction: better inputs, more machine-side reasoning, and clearer guidance. The intended result is less work for the person without giving up the safety backbone of AID.

Automated insulin delivery works by turning clean numbers—carbs, glucose—into insulin, and years of research have refined the math that connects them. That math is real, and it's genuinely hard-won.

But an algorithm can only act on what it's given. Around the control law sits another layer that shapes every result: how much the system knows about you, how it handles the messy and uncertain inputs real life produces, and how much of the relentless daily work it takes off your hands. That layer is where Dossi puts its attention.

The frontier isn't a smarter control law. It's a system that knows more and does more—so you do less.

That is the bet Dossi is built on. And the two payoffs people usually treat as a trade-off—better numbers and a lighter load—fall out of the same design choice. Richer, more forgiving handling makes for better dosing directly; a system that asks less of you gets used more consistently and fought less, which is where real-world control actually comes from.

It shows up in three places.

1. Better inputs

Take what's human, not what's precise.

Every other system needs a number for every meal, and the dose is only as good as your estimate. Dossi lets you log a meal by photo, a sentence, or a conversation, and carries carbs as a low/mid/high range instead of a false exact. When you're unsure, that uncertainty flows into the dose—it acts more cautiously and won't mislearn a rough guess as an error. No shipping AID, commercial or DIY, works this way.

2. Better personalization

It learns you from what actually happens.

Dossi models eight kinds of meal, each with its own absorption curve refined to your body, and learns your insulin sensitivity, carb ratios, and dawn pattern from your clean days. It eases insulin on high-movement days and detects exercise live—no mode to remember—and carries the long sensitivity tail after a workout. After a low, it holds off re-dosing until you're genuinely stable and leaves that rebound out of learning, so a juice box is never mistaken for insulin resistance. Context that is not safe to turn into a generic learned factor stays separate from dosing.

3. Better experience

Ask less, forgive more.

The everyday micro-decisions become the machine's job, not yours. You stay in command—when you choose to bolus, Dossi warns rather than blocks; your call is yours. And none of the “do more” touches the safety net: suspend-on-low, forecast-gated resume, hypo protection, and dose caps stay enforced no matter what.

Dossi isn't trying to replace the category with something unfamiliar. It keeps the closed-loop backbone that already works and focuses on what the algorithm can't do alone—the information you give it, the inputs it forgives, and the daily load it can lift. That is why it exists.