Fully-closed-loop algorithms have arrived. Now we all have a hard choice to make.

The real question isn’t whether we can remove carb counting. It’s what results we’re willing to accept when we do.

For decades, people with type 1 diabetes have heard about the promise of an “artificial pancreas”: a system that automatically manages glucose levels without needing carb counting or meal announcements.

Hybrid closed-loop systems have already brought us much closer to that goal. Pumps like MiniMed 780G, Omnipod 5, Tandem Control-IQ and CamAPS FX automatically adjust insulin throughout the day based on CGM readings.

But they still rely on one key piece of human input: telling the system when you eat, and how much.

Now the industry is racing toward the next milestone: fully-closed-loop insulin delivery, where the system operates with no user input at all.

In March 2026, we may have seen the first major steps in that race, with announcements from both CamDiab and MiniMed.

But the announcement also raises a bigger question that the diabetes community may not have really answered yet.

What level of glucose outcomes is actually acceptable when there is zero user input?

CamDiab Announces a “Fully-Closed Loop” Feature

Last week, Cambridge-based company CamDiab announced a new, fully-closed-loop mode for its CamAPS FX algorithm (used by Ypsomed pumps) that will soon be submitted to regulators. (Keep in mind we probably won't have this in Australia for 1–3 years).

The system is designed to run without carbohydrate counting or meal announcements.

But the company’s own press release contains an important qualifier:

“The feature offers an option for people who may benefit from reducing the cognitive and emotional load of diabetes management at key moments in their lifelong journey with type 1 diabetes… The feature is designed to complement, not replace, an individual's existing diabetes management plan.”

The announcement suggests the feature may be most useful during "periods of stress, travel, busy work or school commitments, or other life events". In other words, it may be something people switch on and off, rather than something they use permanently.

That framing is interesting. Because if a system performs better when users return to carb counting, then the “fully-closed-loop” mode may simply be trading glucose outcomes for convenience. And is that really a fully-closed loop?

Is this any different to Tesla's Auto Pilot feature? The name implies that the car drives itself but in reality the driver has to be ready to take the wheel at any point.

How small does the difference need to be for a "fully-closed loop" to earn the title? Up to 10% less time in range? 5%? No difference at all?

MiniMed Enters the Chat

Only days later, at the Advanced Technologies & Treatment for Diabetes (ATTD) conference in Barcelona, MiniMed revealed feasibility study results for its own fully-closed-loop algorithm, which they're calling Vivera. We have a lot more technical information about this new algorithm, including the time-in-range results from their trial.

Here are the highlights:

• There's no distinction between hybrid and fully-closed loop; users can either enter carbs, or not, and they can do that at any time. If the pump has the carb data, it will take that into account; if not, it will detect the blood glucose rise and adjust as best it can.

• There will be no more SmartGuard exits (cue the applause from current users!)

• Blood glucose targets have been expanded, with a new option for 5mmol/L, as well as 5.5, 6.1, 6.7, 7.2 and 7.8. The temp target (for exercise) will be 8.3mmol/L.

During the study of 14 adults using this system, they were able to achieve an average of 74% time in range. The 13 children in the study achieved an average of 65% time in range.

In terms of release date, there's a pivotal trial to be done next, and then of course the approval by various regulatory authorities. So again, probably a few years before we see it in Australia. MiniMed have said that Vivera will launch with MiniMed's new 8-series pump (Flex), which we know is currently with the FDA for approval.

What Does “Good Enough” Look Like With Zero Input?

International diabetes guidelines generally recommend aiming for at least 70% time in range to reduce the risk of long-term complications.

That might sound like a reasonable minimum target for a fully-closed-loop system.

But is it enough?

Some people using existing hybrid-closed-loop systems are able to get their time in range to 80% or above¹ when carb counting is done well. So if removing carb counting drops outcomes back toward 70%, are we really moving forward? Or are we simply trading glucose control for convenience?

This is the debate the diabetes community hasn’t had yet.

The key question isn’t whether fully-closed-loop systems are possible. It’s "what level of performance we are willing to accept in exchange for a reduced mental load?".

Where Each Company Stands

Several insulin pump companies have hinted at their plans for fully-closed-loop systems, though most stop short of defining exactly what that means in practice.

MiniMed’s recent announcement is the most explicit and detailed right now, with actual study results to show an expectation of time in range.

CamDiab hasn't released any data, but their press release wording suggests the system may not always deliver the same glucose outcomes as when meals are announced.

Other companies have been more cautious in their language.

Tandem Diabetes Care has said it is working on improving how algorithms handle “unannounced meals”.

Insulet’s public roadmap places fully-closed-loop insulin delivery after 2028, suggesting the company believes significant technological progress is still needed before such systems are ready.

It will be interesting to see how each company frames the feature—and what outcomes they claim they can deliver—when they announce their respective versions.

Why This Problem Is So Hard

Even with perfect algorithms, insulin has a fundamental limitation. It works slowly. Even the fastest insulins:

• begin working after about 10–15 minutes

• reach peak effect around 60–90 minutes

Meals can raise glucose much faster than that. So without advance warning, an algorithm is always reacting after glucose has already started rising. This delay is one of the biggest reasons fully-closed-loop systems remain such a difficult problem to solve.

The Bigger Conversation We Need to Have

The race to fully-closed-loop insulin delivery is clearly underway. But before these systems arrive, the diabetes community may need to answer a simple question:

What results are good enough when the user does nothing?

Is 70% time in range acceptable if the system requires zero carb counting?

Should fully-closed-loop systems aim for 80% or higher to match what many hybrid systems already achieve?

Or is the real goal something else entirely: reducing the mental load of diabetes, even if glucose outcomes aren’t perfect?

You can read the complete CamDiab press release here.

1. van den Heuvel T et al. Diabetes Technol Ther. 2024;26(10)