Skip to content

About

Daymark: a private, local-first task scheduler and day planner. Native iOS app in progress on a shared Rust core, with macOS, Windows, Linux, and Android coming. Showcase only, not the source.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Repository files navigation

Daymark

Daymark is a private, local-first task scheduler and day planner. You type what you need to get done, it turns that into dated, timed tasks on a calendar, and the whole plan stays on your own device. No account, no telemetry.

It started as a web app. Right now I'm rebuilding it as native apps, one per platform, all running on one shared Rust core. iOS is first and it's the one I'm actively building.

Live page: https://josephcintron.github.io/Daymark_Showcase/

This repo is a showcase, not the source. Daymark is proprietary. What's here is the write-up, screenshots, a few simplified code excerpts, and one small runnable demo.

The iOS app

Written in SwiftUI (iOS 17+) on top of the Rust core. These screenshots are from the latest iOS build (native 0.1.39).

Today Calendar Tasks
Today. Opens on what's left and what's next. Calendar. Month and day views. Tasks. Subtasks, filters, and a week at a time.
Draft with AI Ask Daymark Edit Dashboard
Draft with AI. Apple's on-device model drafts the tasks and I review them before anything gets added. Ask Daymark. Plain-language changes show up as a preview first. Dashboard editing. Hold, drag, and stack widgets like the Home Screen.
Speak a task Dark mode Welcome
Speak a task. Say it and Daymark reads the date and time. Dark mode. Light, dark, or follow the system. Welcome. Pick a direction and get a starting plan you can edit.

What works on iPhone so far:

  • Quick add that reads dates and times as you type ("Dentist Thu 3pm", "gym saturday 9am and groceries sunday")
  • Today, Tasks, Calendar (month and day), and Search
  • Every change goes through one set of validated operations, and all of it can be undone
  • Draft with AI and Ask Daymark run on Apple's on-device model, so nothing leaves the phone
  • Optional AI details (a short description and a time estimate), off until you turn it on
  • Time estimates that learn from your own tasks
  • Home Screen widgets, Siri, a Spotify widget, and Settings laid out like iOS Settings
  • A sample plan to try it out, and it opens backups from the web app

Coming soon

Platform Native app What runs today
iPhone, iPad SwiftUI on the Rust core In progress
macOS Same SwiftUI codebase, Mac target Web app in a Tauri shell
Windows C# and WinUI 3 on the Rust core Web app in a Tauri shell
Linux Rust with GTK4 and libadwaita Web app in a Tauri shell
Android Coming later Web app in a Capacitor wrapper

Until each native app is ready, the web app keeps shipping. On desktop it's packaged with Tauri, so Windows, macOS, and Linux still get an installable app in the meantime.

The web and desktop app

This is the desktop build as it is right now.

Dashboard

Calendar Dark mode
Month calendar with drag scheduling. Dark mode.

Why native

I want direct access to on-device ML and the GPU on each platform: Core ML and Apple's on-device model on Apple devices, Windows ML and DirectML on Windows. A wrapper can't get to that cleanly, so I'm doing the full rewrite.

What keeps that manageable is one shared core. All the logic lives once, in Rust: the plan format, the operations and Undo, reading dates and times, the calendar math, search, repeats, and estimates. Each app reaches it through UniFFI, and the core is tested against the web app's own results so every platform behaves the same. Each app only writes its own interface and its own ML.

More detail in ARCHITECTURE.md.

How I think about it

  • Today first. It opens on today: what's left and what's next.
  • Ask, don't guess. If you don't say when something happens, it asks. If a date is ambiguous it leaves it alone.
  • Your word wins. A time you typed is yours. An AI estimate stays marked as a guess and never overwrites what you set.
  • One way to change the plan. UI, keyboard, voice, or AI, it all goes through the same validated operations. A failed change changes nothing.
  • Nothing gets lost. Saves are atomic, unreadable data is set aside instead of overwritten, and backups move between every app unchanged.
  • Private by default. No account, no telemetry. AI is opt-in and sends the minimum.

Where it started

Daymark grew out of a job-search tool I built: a service that pulled postings from public ATS APIs (Greenhouse, Lever, Ashby), filtered them with regex, and scored the rest with a local model into structured JSON. That part is on hold while I focus on scheduling. The constrained-scoring demo and the ML methodology page come from that work.

What's in this repo

License

Proprietary. © 2026 Joseph Cintron. All rights reserved. See LICENSE. The snippets and demo are here for review only.

About me

I'm Joseph Cintron, a Computer Science senior at FIU with an AI/ML concentration. I've spent about three years building automation and IT systems for an accounting firm. I design and build Daymark myself.

Reach me on GitHub: @JosephCintron

About

Daymark: a private, local-first task scheduler and day planner. Native iOS app in progress on a shared Rust core, with macOS, Windows, Linux, and Android coming. Showcase only, not the source.

Topics

Resources

Stars

1 star

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages