A prep task is an auto-generated pre-event checklist — the small actions you need to take before an event so you walk in ready, not scrambling. In Composed, the generate-checklist edge function reads the event’s title, category, date, and days-until, then writes a 3-5 item list with informational tips. Apple Reminders has no auto-generation. Todoist supports manual templates only. Things 3 and Sunsama have neither. Prep tasks fire at the action layer of Composed’s three-layer notification model.
What is a prep task — the actual definition?
A prep task is a to-do item that lives inside a specific calendar event and surfaces in the preparation window before that event fires. In Composed, prep tasks are stored on a Follow-Up Anchor record shared between the event in the calendar and the todo drawer, so the same checklist is reachable from either place.
The distinction matters. A regular to-do is a flat backlog you manage by hand — write it, remember it, archive it. A prep task is event-scoped. It exists because the event exists. It surfaces because the event is approaching. It retires because the event happened. You are not the bookkeeper.
The clearest way to see the difference is to picture a dentist appointment on Tuesday at 2pm. Apple Reminders gives you one notification 30 minutes before. Composed gives you a prep checklist days earlier: bring your insurance card, write down questions about the sensitivity in your left molar, brush before you go, leave with an 18-minute buffer because the parking lot fills up at lunch. The event is the same. What surrounds the event is completely different.
A reminder that you have somewhere to be is not the same as a reminder to prepare. Most planning apps only do the first part. Prep tasks do the second.
For a fuller definition of how this term fits into the Composed vocabulary, the calendar prep tasks glossary entry walks through the data model in more detail.
How does Composed’s AI generate prep tasks automatically?
When you create an event in Composed — by voice-to-calendar, by screenshot, or by typing — the app sends four pieces of context to its generate-checklist edge function: the event title, the category (appointment, travel, social, deadline, wake alarm), the event date, and the number of days until it fires. The function calls Claude through Anthropic’s API and returns 3-5 prep tasks plus optional informational tips, all tagged with relative timing.
Three details about that pipeline matter for understanding what you actually get.
Context-aware, not template-driven. A dentist appointment doesn’t get the same list as a job interview. “Dentist Thursday at 2pm” returns bring insurance card, write down questions for the hygienist, brush before the appointment. “Annual physical Friday at 9am” returns fast for 12 hours if labs are scheduled, bring a list of current medications, write down anything you want to ask the doctor. The category and title together change what the AI considers relevant.
Timing is part of the task. Composed doesn’t just dump the list at the moment of event creation. Each task carries a relative-timing hint — “two days before,” “the night before,” “the morning of” — and the action-layer notifications fire at those moments instead of all at once.
Editable, never imposed. Every AI-generated task has a checkbox. You can remove tasks that don’t apply, add ones the AI missed, and adjust the timing. Composed suggests; you decide. There is no penalty for editing — the readiness score recalculates against your final list, not the original one.
For events created from a screenshot of a flight confirmation, the AI prep tasks feature also pulls flight context (international vs domestic, departure airport, arrival airport) into the same generate-checklist call, so the resulting checklist includes the passport-expiration check on international itineraries and the airport-buffer suggestion (120 minutes domestic, 180 international) on every flight.
What is the difference between a prep task and a regular reminder?
A reminder tells you when. A prep task tells you what to do before that when. Apple Reminders, Todoist, and TickTick all model “remind me about X at Y o’clock.” Composed models “remind me about X, and surface the five things I need to do beforehand at the right moments.”
The distinction shows up in the data model. A reminder is a single notification scheduled for a single moment. A prep task is a checkbox bound to a Follow-Up Anchor that’s bound to an event — meaning the task knows what event it belongs to, the event knows what tasks it has, and the readiness score can compute how many are done.
It also shows up in the behavior over time. Once an event passes, its prep tasks retire with it. They don’t linger in your todo backlog as “old, unfinished items” the way a Todoist task would if you forgot to mark it done. The event is the lifecycle. The tasks ride on the event.
The deeper reason this matters is documented in our why reminders don’t work post, but the short version is: single-trigger notifications fire once and go quiet, which is fine for “the dentist is in 30 minutes” but useless for “you should write down your questions tonight.” Prep tasks fill the part of the timeline reminders ignore.
When does a prep task fire — and when does it get suppressed?
Prep tasks surface in Composed’s action-layer notifications, the middle tier of the three-layer reminder model. Action-layer fires inside seven days of the event. Awareness fires beyond seven days. Urgency fires inside twenty-four hours and never gets suppressed.
The action layer has one specific suppression rule: when your readiness score for the event reaches 95%, action notifications stop. You’ve done the prep; you don’t need more pings. The math is simple — readiness is the percentage of prep tasks you’ve completed, weighted by their importance. Five tasks done out of five is 100%. Three out of five with all critical items checked is around 70%. The 95% threshold is set so that an event with one or two non-critical leftover tasks (the “bring a backup pen” tier) doesn’t keep nagging you when you’re substantively ready.
There’s also a 90-minute cooldown after task completion. Mark a prep task done, and Composed suppresses action-layer pings for that event for the next 90 minutes — so you don’t get a “remember to bring your insurance card” nudge 60 seconds after you ticked it off.
Urgency-layer notifications — the “leave in 12 minutes” departure ping, the boarding-now flight notification — are never gated by readiness. Those run on time and travel math, not preparation status. The reasoning is that “leave by” is a different question from “ready to go,” and you need both signals.
The full architecture is documented in Composed’s notification rules, but the practical takeaway is this: prep tasks nudge gently as long as there’s room to act, and they go quiet the moment you’ve actually prepared. They don’t repeat after midnight. They don’t fire during quiet hours (10pm to 7am) unless they’re urgency-tier. They respect that you already know.
What kinds of events get the best AI prep tasks?
The events that benefit most from AI prep are the ones with hidden logistics — the things that look simple on the calendar but carry a quiet ladder of preparation underneath. A few categories where Composed’s generate-checklist consistently writes useful lists:
Medical appointments. Dentist, annual physical, specialist visit, dermatology check. The AI knows to include insurance-card reminders, medication-list prep, and “write down what you want to ask” — the items that always feel obvious in hindsight and always get forgotten in advance.
Flights. When you screenshot a flight confirmation into Composed, Claude Vision extracts the IATA airport codes, and the prep checklist pulls in the international/domestic context. International flights get the passport-six-month-validity check. Every flight gets the airport buffer math (120 minutes domestic, 180 international, layover-aware on connections). The five flight-specific alerts — check-in at 24 hours, summary at 4 hours, boarding, gate close, layover — fire on top of the prep list.
Events involving other people. Dinner parties where you’re bringing wine. School pickup with a teacher conference attached. A birthday party where you RSVP’d “yes plus one.” The AI doesn’t always know what gift to buy, but it does know to remind you a gift is part of the event.
First-time visits to a new place. A new doctor’s office, a new client meeting, a venue you’ve never been to. The prep list usually includes “look up parking,” “confirm the address,” and “leave an extra 15 minutes” — the small frictions you only learn the hard way.
Recurring family logistics. Soccer tournaments, school picture day, dance recitals. The AI catches the permission-slip, the cash-only concession, the “wear the right color shirt” — items parents know to think about and still forget half the time.
A real-world example: the Composed in the Wild dentist follow-up post walks through what happened when Composed’s founder, Jesse Meria, scheduled a follow-up at the front desk by voice. Ten seconds of speaking. The event landed on the calendar with the address attached, and the prep checklist appeared before the receptionist had finished typing — confirm the appointment the morning of, check the insurance file, plan the drive. That sequence is the standard Composed flow, not a one-off.
The events that don’t benefit much are the simple ones: a phone call from your couch, a coffee with a friend who doesn’t care about timing, a weekly recurring meeting where the prep is always identical and you’ve already automated it in your head. Composed will still generate a list for those, but you’ll usually edit it down to one or two items.
Can prep tasks work for ADHD brains?
Yes — and the architecture of prep tasks specifically is one of Composed’s strongest fits for ADHD planning, because the failure mode prep tasks address is exactly the failure mode time blindness creates.
Russell Barkley’s research on ADHD and executive function (ADHD and the Nature of Self-Control, 1997, and subsequent papers through 2023) identifies time blindness as a structural deficit, not a discipline failure. The brain doesn’t represent future time the way it represents present time, so prep work — which exists in future time — never quite feels real until the future arrives. This is why ADHD adults can know an event is coming, know what it requires, and still arrive unprepared.
Prep tasks bridge that gap in three ways the ADHD planning guide goes deeper on:
The thinking happens once, not many times. When you create the event, the AI does the “what does this require” pass that an ADHD brain would otherwise have to do over and over in scattered moments. The prep list is the result of that thinking, captured before you lose it.
Tasks surface when they can be acted on. Action-layer notifications fire two days before, the night before, the morning of — not at the moment of event creation when the future feels distant. By the time the task surfaces, the event is close enough that “future” has compressed into “soon.”
The yellow card, not the red badge. Composed’s deadline cards are yellow (#FDE047), never red. The shame-coded red-badge pattern that triggers avoidance in ADHD brains is absent by design. A prep task that’s still incomplete the morning of the event shows up as a calm yellow nudge, not a jarring screen.
The combination — AI-generated lists that capture the thinking up front, action-layer nudges that fire at the actionable moment, and calm visual language throughout — addresses the documented failure mode rather than fighting it.
Where prep tasks sit in the bigger picture
The reason prep tasks are a distinct concept in Composed and not in Apple Reminders, Todoist, Things 3, or Sunsama is that prep tasks require three things to coexist: an event-scoped data model, an AI capable of inferring relevant preparation from event context, and a three-tier notification system that knows when to nudge and when to stay quiet.
Apple Reminders has the notifications but no AI and no event scope — every Reminder is a top-level item. Todoist has templates but no auto-generation from event context. Things 3 has elegant projects but no AI inference. Sunsama has daily-planning rituals but no preparation-window logic. None of those are bad apps; they’re solving different problems. Composed is the iOS planner that treats “the work before the event” as a first-class concept with its own data model, AI, and notification tier. The planner alternatives hub compares Composed against each of those apps in more detail.
A prep task is not the answer to every planning question. If your goal is to manage a project with twelve dependencies and four assignees, you want Things 3 or Notion. If your goal is to capture a stream of thoughts as flat tasks, Apple Reminders is fine. But if your goal is to walk into events ready instead of scrambling — the dentist appointment with the insurance card in your wallet, the flight with the boarding pass loaded and the airport buffer planned, the dinner party with the wine actually in hand — the prep task is the primitive built for that.
You can start the habit today without any app at all. Pick one event this week. Spend thirty seconds asking what it actually requires. Write three answers somewhere you’ll see them — what to have, what to do, what to confirm. That’s the manual version, and it works. The reason Composed automates it is that the manual version requires a context-switch most people skip when they’re adding events quickly. The AI removes the friction; the underlying habit is the same.
Frequently asked questions
What does Composed's generate-checklist edge function actually return?
The generate-checklist edge function returns 3-5 prep tasks per event, each with a description, an optional informational tip, and relative timing (the night before, the morning of, two days out). For flight events it also pulls in passport-validity checks and airport-buffer recommendations based on international or domestic context. The function reads four required inputs — title, category, date, and days_until — plus optional location, travel_time, and flight_context.
Do prep tasks work without an internet connection on iPhone?
Composed requires an internet connection to generate prep tasks because the generate-checklist function runs in the cloud, not on-device. Once tasks are generated and saved to the event, they sync to local cache and remain visible and checkable offline. New events created offline queue their AI generation request and complete it when iPhone reconnects.
Can I add my own prep tasks instead of using the AI-generated ones?
Yes. Every prep task in Composed is editable. You can add tasks the AI missed, remove tasks that don't apply to your situation, and adjust the relative timing of each item. The readiness score recalculates against your final list, and the action-layer notifications fire on your edited timing — not the AI's original suggestion.
Why do prep tasks disappear after the event happens?
Prep tasks are scoped to their parent event, so when the event date passes, the task list archives with the event instead of lingering in the todo backlog. This is the deliberate behavior — a task to bring your insurance card to a dentist appointment last Tuesday has no useful life past Tuesday. The Follow-Up Anchor architecture in Composed stores the tasks on a shared record that retires with the event.
How is a prep task different from a subtask?
A subtask in Apple Reminders or Todoist is a child item under a parent task, used to break a larger task into smaller steps. A prep task in Composed is a checklist item bound to a calendar event, surfacing in the preparation window before that event. The structural difference is that prep tasks have event-scoped timing and retire with the event, while subtasks live inside a task hierarchy with no event relationship.
