Personal Instruments
Near launch

SayList

An iPhone voice-capture instrument that turns natural speech into a reviewable task list, then lets the user edit, share, or create Apple Reminders.

Category: Personal Instruments
Platform: iPhone · iOS 26+
Status: Near launch

What if speaking several tasks once were enough to create a list you can review?

Visible LISTEN and STOP capture, Apple Speech, layered Apple Intelligence with deterministic fallback, local task storage, review, sharing, Reminders, widgets, and Live Activities keep the flow direct.

User value

Speak the list once. Review every task before it acts.

Capture errands, follow-ups, appointments, shopping reminders, and quick obligations in ordinary speech without organizing them before you begin.

Capture several tasks naturally

Tap one prominent control, speak with ordinary conjunctions, dates, places, and context, then stop when the thought is complete.

Keep extraction available

Use validated Apple Intelligence output on supported, ready devices while retaining a deterministic extractor when model assistance is unavailable or disallowed.

Review before external action

Inspect, edit, complete, delete, share, or create Apple Reminders only after the generated task records are visible.

SayList iPhone idle screen with one large blue LISTEN control centered on a black background. — **Current capture surface**Current simulator evidence showing the single-action idle state.

Features

A capture-first path from speech to tasks.

The current v1 implementation keeps recording simple and moves correction, sharing, and Reminders into an explicit review surface.

01
Visible LISTEN and STOP capture
Begin only from an explicit foreground action, show recording state and live transcript feedback, and stop through the same primary control.
02
Layered task extraction
Use rules for provisional feedback and fallback, then attempt validated Foundation Models extraction for stable transcripts when the device and access policy permit it.
03
Review and supported editing
Complete or delete tasks and edit the title, point due date, and place while keeping inferred date windows and context visible.
04
Local task persistence
Store task records in a local JSON file, suppress duplicates, and show calm feedback when loading, saving, or insertion does not succeed.
05
Sharing and Apple Reminders
Share plain text or a text file and create one or several Reminders only after an explicit user action and system permission.
06
Quick entry and live status
Use a Home Screen widget or URL route to start one visible session, then follow recording, processing, ready, or error states through a Live Activity.

How it works

Listen, extract, review, act.

SayList separates the time-sensitive capture moment from the slower responsibility of checking generated tasks.

01
Tap LISTEN
Grant microphone and speech permissions when needed and begin one visibly represented recording session.
02
Speak naturally
Describe several obligations in one passage while Apple Speech produces a transcript and rules provide provisional task feedback.
03
Tap STOP
Finalize transcription and send the stable text into the coordinated extraction path.
04
Extract with fallback
Validate model-assisted candidates when Apple Intelligence is eligible, or use the rule-based extractor after unavailability, timeout, failure, or invalid output.
05
Review the task records
Inspect wording, dates, places, and context, then edit supported fields, complete items, or delete mistakes.
06
Choose the destination
Keep tasks locally, share them as text or a file, or create Apple Reminders through an explicit handoff.

In practice

One action, then a reviewable list.

SayList recording screen with a large red STOP control, two-task capture indicator, and a short synthetic transcript. — Current deterministic simulator recording state.

Capture in progress

The control changes to STOP while the interface shows a current-session task count and the working transcript.

SayList idle screen with a blue LISTEN control and a Review tasks status showing two new items. — Current simulator ready state after extraction.

Tasks ready

After processing, the main surface returns to LISTEN and exposes the number of new tasks waiting for review.

SayList Captured Tasks sheet with three synthetic tasks, dates, a place, contextual timing, and a Create Reminders action. — Current review sheet using fictional task data.

Review before action

The secondary sheet shows task titles, dates, places, and context before sharing or Reminders creation.

System surfaces

Recording state and optional intelligence remain visible.

These current simulator captures document implemented surfaces. Final hardware checks, StoreKit behavior, and marketing approval remain pending.

iPhone Lock Screen with a SayList recording Live Activity showing 28 seconds elapsed, two tasks captured, and a red Stop button. — Current simulator Live Activity evidence.

Lock Screen Live Activity

A user-started recording can show elapsed time, current-session feedback, and a Stop action while brief background continuation remains bounded to three minutes.

SayList Plus screen explaining the fifteen included Apple Intelligence captures, one-time unlock, and continued standard extraction. — Current simulator Plus explanation without an asserted storefront price.

SayList Plus boundary

Fifteen qualifying Apple Intelligence captures are included before a one-time Plus unlock, while standard extraction and the core workflow remain available.

Product concept

Capture and review are different moments.

Speaking is fastest when obligations are still unstructured. Review is where generated wording, dates, places, and context become accountable.

Ordinary speech comes first

The user should not need to dictate perfect commands or navigate a task hierarchy before preserving what needs to be done.

Intelligence is layered, not exclusive

Model-assisted interpretation can improve natural phrasing on supported devices, while deterministic rules keep the basic product available and testable.

External action follows review

Sharing and Reminders belong after the person can inspect and correct the individual tasks produced from the transcript.

Architecture

A local task pipeline with Apple service boundaries.

The iPhone app owns the visible recording lifecycle, extraction coordination, task review, local persistence, and handoffs. Apple frameworks provide speech, optional model execution, Reminders, StoreKit, widgets, and Live Activities.

01Explicit user speech capture
02Apple Speech transcript
03Layered extraction coordinator
04Validated TaskItem records
05Local JSON persistence and review
06Optional share or Apple Reminders

Recording stays visible and bounded

Capture begins only from a user action, remains represented in the app or Live Activity, and can continue in the background for at most three minutes after a brief switch.

Rules remain a dependable fallback

The deterministic extractor handles provisional work and stable fallback when Apple Intelligence is unavailable, not allowed, times out, fails, or yields no valid tasks.

The task store remains local

SayList uses local JSON and contains no app-owned backend, account, analytics pipeline, cloud sync, or cloud language-model endpoint.

Local-first is not an offline guarantee

New voice capture still depends on Apple Speech availability, and Apple Intelligence behavior depends on device support, settings, language, and model readiness.

Technology

Native capture with layered extraction.

Application

Swift
SwiftUI
iPhone
iOS 26+

Capture and surfaces

AVFoundation
Apple Speech
WidgetKit
ActivityKit
URL routes

Extraction

Foundation Models
Structured validation
Deterministic TaskExtractor
Fallback coordinator

Data and handoff

Local JSON
EventKit
System share sheet
StoreKit 2
UserDefaults allowance

Current boundary

No app backend
No accounts
No analytics
No cloud sync
No cloud LLM

Development

The v1 product is mature. External release gates remain open.

The complete Record, Stop, extract, review, edit, share, and Reminders path is implemented with local persistence, duplicate suppression, widgets, deep links, limited background continuation, Live Activities, Plus entitlement logic, and deterministic evidence tooling. Physical-device sanity, live StoreKit behavior, signing, distribution, required URLs, privacy review, and current marketing assets remain incomplete or unconfirmed.

Stage: Near launch
Platform: iPhone · iOS 26+
Automated release gate: 291 tests · 0 failures
Public availability: Not confirmed

Implemented

Visible LISTEN and STOP capture, live transcript feedback, extraction progress, and review-ready states
Apple Speech transcription, validated Foundation Models extraction, deterministic provisional processing and fallback
Editable local task records with due dates, date windows, places, context, duplicate suppression, and persistence warnings
Plain-text and text-file sharing plus single-task and batch Apple Reminders handoff
Widget and URL capture entry, ActivityKit recording state, Stop and Review actions, and bounded background continuation
SayList Plus allowance, purchase and restore interface, deterministic scenarios, release builds, accessibility tests, visual evidence, and privacy audits

Before launch

Complete the physical-iPhone speech, widget, Lock Screen, interruption, three-minute stop, permission, and Reminders matrix. Validate App Store Connect product discovery, purchase, restore, refund, and revocation, finish signing and upload, supply support and privacy URLs, complete the privacy questionnaire, and regenerate and approve current-interface screenshots and preview media.

SayList lab notes

The project, over time.

Notes about the thinking, architecture, development, and decisions behind this work.

Speak the list once. Review every task before it acts.

Capture several tasks naturally

Keep extraction available

Review before external action

A capture-first path from speech to tasks.

Visible LISTEN and STOP capture

Layered task extraction

Review and supported editing

Local task persistence

Sharing and Apple Reminders

Quick entry and live status