Tasky - Your Productivity Pal

Tasky: The smart to-do app. Type tasks in natural language, and our NLP schedules them instantly. Stop managing, start doing. Install on all devices as a offline-first PWA.

Comment

About Tasky

Tasky is a smart, minimalist task manager that turns natural language into structured, actionable plans. Type what’s in your head - “Email Sam by 5pm urgent #work” - and Tasky parses the time, priority, tags, and context instantly so you can focus on doing, not managing.

What inspired us

  • We kept abandoning to-do apps because they added friction. Lists don’t think; they just store. I wanted input to feel as natural as thought, and output to feel like momentum.
  • We love the idea that small wins compound. Streaks, focus sessions, and tiny insights can change how we work — without adding complexity.

What we built

  • Natural language task input: Understands dates, times, priorities, and tags from plain English.
  • Smart suggestions: Context-aware prompts (e.g., “tonight”, “tomorrow morning”) based on time-of-day and task type.
  • Focus sessions: A distraction-free mode with timers that feeds into insights and streaks.
  • Streaks and insights: Lightweight analytics that turn consistency into motivation.
  • Sections and notes: Organize tasks into Today, Tomorrow, Upcoming, with optional notes.
  • Secure sync: User-specific data with Firestore rules and Auth; designed for offline resilience.

How we built it

  • Frontend: TypeScript + modern React patterns for a fast, ergonomic UI.
  • Smart parsing:
    • chrono-node for robust date/time extraction.
    • Custom NLP for priority inference, tag extraction, title cleanup, and suggestion generation.
    • A confidence model to gauge parse quality and adjust UX accordingly.
  • Data & auth: Firebase Auth + Cloud Firestore with user-scoped security rules.
  • Resilience: IndexedDB-backed persistence to keep things responsive offline.

Challenges we faced

  • Ambiguous language: “Call Alex next week” vs “Call Alex by 5” required careful heuristics to avoid over-confident guesses. I default to sane times (e.g., 11:59 PM) only when the user didn’t specify a time.
  • Priority inference: Natural phrasing (“no rush”, “by EOD”, “!”) needed contextual, negation-aware rules to avoid false highs and lows.
  • Security rules: Ensuring every read/write is scoped to the authenticated user across multiple collections (tasks, insights, preferences) while keeping DX smooth.
  • Offline behavior: Balancing IndexedDB persistence with real-time updates and avoiding confusing UI states.

What we learned

  • Small UX details compound: fast debounced parsing, three great suggestions, and predictable defaults dramatically reduce friction.
  • Hybrid approaches beat pure ML here: deterministic parsing + targeted heuristics gives reliability and transparency.
  • Invest early in Auth + Rules: well-structured, user-scoped Firestore rules pay off in confidence and maintainability.
  • Opinionated “focus first” design helps people actually get things done.

What’s next

  • Smarter daily planning: one-tap realistic plans that fit your pattern of work.
  • Deeper insights: trendlines for procrastination hotspots, best times to work, and task types that bog you down.
  • Calendar bridges: optional two-way sync to make time-blocking effortless.
  • Lightweight collaboration: shared tasks with the same natural input flow.

If you want a task manager that thinks with you - not at you - try Tasky.

Built With

Share this project:

Updates