The most advanced open-source AID system for iPhone — built on the oref1 algorithm with Super Micro Boluses, Unannounced Meal detection, and Dynamic settings that adapt to your insulin sensitivity in real time. This is what Arden uses.
Device specifications change frequently — always verify current information directly with the manufacturer before making any decisions. Full disclaimer.
Curious what Trio is? Start here.
Trio is not a commercial product and is not FDA-cleared. It is open-source software built and maintained by the diabetes community. You build it yourself, you maintain it yourself, and you use it at your own risk — with community support. Read the full TrioDocs documentation before starting.
This is the system Arden Benner uses. Scott has covered the iAPS/Trio lineage on the Juicebox Podcast — most directly in Episode #1243: DIY Insulin Algorithm — iAPS with Deniz from Germany, who later moved to Trio. The full Algorithm Pumping series is at juiceboxpodcast.com → Series → Algorithm Pumping.
If Loop is automated insulin delivery for people who want a reliable, well-documented system, Trio is for people who've been looping and want more — more responsiveness, more automation, more adaptability. It runs the oref1 algorithm, which was designed from the start to handle the things Loop deliberately doesn't try to do: meals you didn't announce, sensitivity that shifts throughout the day, and corrections that act through rapid micro-boluses rather than slower basal adjustments.
When glucose is rising, Trio delivers small automatic correction boluses every 5 minutes — front-loading insulin rather than waiting for basal adjustment to take effect. This produces faster corrections and better post-meal control than pure basal adjustment.
Trio detects an unexpected glucose rise and treats it as if you ate without announcing the meal — dosing insulin automatically via SMB even when no carbs were entered. Loop requires you to either bolus or enter "fake carbs" for this scenario. Trio handles it.
Trio adjusts your insulin sensitivity factor (ISF) and carb ratio (CR) in real time based on your current glucose level and trend. If glucose is running high, it becomes more aggressive. If running low, more conservative. Loop uses fixed values you program.
Autosens continuously analyzes 24–48 hours of data and adjusts algorithm behavior if you're more sensitive or resistant than usual. Autotune reviews 7+ days and recommends setting changes. Neither is available in commercial AID systems.
Arden's using iAPS now. She used Loop before that, she's used Omnipod 5 — we've used a lot of different algorithms. But it took me back to managing Arden before you know, an algorithm wasn't a thing when Arden was eight years old.
— Scott Benner · Juicebox Podcast Episode #1243 · juiceboxpodcast.com/episodes/jbp1243OpenAPS (2014) → FreeAPS → FreeAPS X → iAPS → Trio. Each fork brought the oref1 algorithm to a new platform or added new capabilities. Trio is the current maintained version for iPhone — it forked from iAPS after a version 3.0.0 disagreement in the development community about code practices and openness. The algorithm underneath is the same one Dana Lewis and Scott Leibrand designed a decade ago. The community around it is what keeps it alive and improving.
Trio users are typically experienced loopers who moved past standard Loop for more performance. They understand their settings, they read the algorithm decisions, and they often manage their diabetes with impressive sophistication. Take their reports seriously.
Dynamic features adapt around your base settings — they don't replace them. Bad basal rates, wrong ISF, or miscalibrated CR will produce bad results regardless of how sophisticated the algorithm is. Your clinical input on base settings is still essential.
The official Trio recommendation: wait 7 days after starting (or switching settings) before enabling Dynamic ISF/CR or Autosens. The algorithm needs data to adapt. Clinically, this patience applies to settings changes too — give changes time before chasing more adjustments.
Most Trio users connect to Nightscout for remote monitoring and data. Ask them to share their Nightscout URL. You'll see CGM data, SMB deliveries, autosens ratios, and algorithm decisions — often more detail than any commercial system provides.
Trio's most important feature for new users: you don't have to turn everything on at once. By default, Trio acts like a regular pump with only occasional temp basal suggestions. You activate capabilities progressively, gaining confidence at each step before moving to the next. This is not how it's always explained, but it's how it should be done.
Activation StepsTrio is installed and connected to your pump and CGM. It behaves like a standard pump — you bolus manually, basal runs as programmed. Trio may occasionally suggest a temp basal adjustment but takes no automatic action. Use this to verify everything is connected and your settings are loaded correctly.
Trio now adjusts basal delivery automatically every 5 minutes based on CGM data and predicted glucose. It will suspend or reduce basal to prevent lows, and increase basal to prevent highs. Essentially the same behavior as standard Loop at this point. Run here for several days before continuing.
Super Micro Boluses activate. Instead of only adjusting basal, Trio now delivers small correction boluses automatically when glucose is predicted to rise. These are typically 0.2–0.5 units delivered every 5 minutes. Post-meal control improves significantly. Max SMB size is configurable — start with conservative settings.
Trio detects unexpected glucose rises — from meals you forgot to bolus for, inaccurate carb estimates, or high-fat meals — and delivers SMBs in response even without a carb entry. This eliminates the "fake carbs" workaround that Loop users learn. Enable UAM and SMBs together; they're designed to work as a pair.
Your insulin sensitivity factor and carb ratio now adjust in real time based on current glucose and trend. When glucose is running high, Trio becomes more aggressive. When running low, more conservative. Wait at least 7 days at Step 3 before enabling this. Requires well-calibrated base settings to function safely.
Autosens analyzes 24–48 hours of data continuously, looking for patterns that suggest you're more or less sensitive to insulin than your settings reflect. It then adjusts basal, ISF, and target glucose to compensate — automatically, within configurable safety limits. Useful for illness, hormonal cycles, high/low activity weeks.
Ask which features they have enabled. A patient on Closed Loop only is in very different territory than one running SMBs + UAM + Dynamic ISF. Calibrate your conversation to their actual configuration.
Nightscout shows SMB deliveries, autosens ratios, and IOB alongside CGM values. High SMB frequency with glucose still running high = I:C ratio probably off, or absorption much faster than expected. Frequent zero temps during day = SMBs are firing correctly.
Ask if they've run Autotune recently and what it suggested. Autotune frequently catches systematic setting errors that aren't obvious from clinic data alone. If Autotune consistently recommends a lower basal overnight, that's clinically actionable information.
Don't try to change this. The 10-hour DIA is load-bearing in the oref1 algorithm's IOB calculations. Attempts to shorten it to 4–6 hours (as in commercial systems) will break the algorithm's safety model. This is not a typo in their settings.
Trio is one of the most sophisticated closed-loop algorithms available to people with diabetes — and it's free. Understanding how oref1 actually works helps you configure it intelligently, interpret what it's doing, and trust it when the numbers look counterintuitive.
oref1 — The Algorithm InsideThe most recent glucose reading and the direction and speed of change — used to estimate where glucose will be in 30–60 minutes.
How much active insulin is currently in the body from all previous deliveries — basal, bolus, and SMBs. The algorithm must never give insulin that, combined with existing IOB, would produce a low. DIA = 10 hours for accurate tail modeling.
Estimated unabsorbed carbohydrates from announced meals. If COB is positive, the algorithm knows glucose may continue rising and can deliver SMBs more aggressively.
The current sensitivity adjustment factor — 1.0 = you're responding normally. Above 1.0 = more sensitive than usual. Below 1.0 = more resistant. This ratio scales basal, ISF, and target glucose proportionally.
oref1 runs five different prediction models simultaneously (standard, insulin only, carbs, unannounced meal, zero temp) and averages them to produce a final prediction — more robust than a single model.
Max Basal, Max IOB, Max SMB Basal Minutes, and Autosens Min/Max all serve as hard stops. The algorithm cannot violate these no matter what its prediction says. They're the safety rails around the algorithm's autonomy.
Source: OpenAPS documentation · TrioDocs · oref1 algorithm specification
The official documentation. New user setup guide, settings reference, FAQ, and transition guides from Loop and commercial systems. The first place to send any new Trio user.
Primary community support. Thousands of active members — many are highly experienced and answer questions with clinical depth. Search before posting; most questions have been answered. Private group requiring approval.
Episode #1243 (iAPS with Deniz) is the most relevant Juicebox content. The Algorithm Pumping series (39 episodes) covers the broader DIY AID landscape and real-world user experience better than any other podcast resource.
Dana Lewis's site — origin of the oref0/oref1 algorithm that Trio implements. Free book "Automated Insulin Delivery" is the clearest explanation of why these algorithms work the way they do. Essential background reading.