ProcrastiNOT

The Problem

Most task managers are either too simple (lacking structure for complex projects) or too heavy (requiring a system to manage the system). AI-enhanced task managers exist but tend to bolt AI onto an existing product as a gimmick rather than designing around it.

The specific pain point: procrastination isn't a willpower problem, it's an information problem. Which task is actually most important right now? What's blocking me? ProcrastiNOT is built around answering those questions intelligently.

Approach

ProcrastiNOT is a native SwiftUI application using SwiftData for persistence, targeting both macOS and iOS as a real App Store product — not a demo or prototype.

Development follows a tiered sprint structure. Tiers completed to date:

• Natural language command bar — TaskCommandParser and TaskCommandExecutor handle freeform input like "add call dentist tomorrow high priority" and translate it to structured task creation
• Focus Mode — AI-powered task selection that surfaces the single most important task based on priority, deadlines, and procrastination history
• Procrastination tracking — records when tasks are deferred, builds a history, and uses it to inform Focus Mode recommendations
• AIInsightsView sidebar — persistent sidebar showing patterns, suggestions, and encouragement based on task history

Tier 6 (iOS) is next, bringing the full feature set to iPhone and iPad with platform-appropriate UI adaptations.

Outcome

Tier 5 complete with all core macOS features functional. The natural language parser handles a wide range of input patterns. Focus Mode is operational. A category filter dropdown bug (auto-advancing on selection) has been identified and queued for the next session.

Stack

Swift and SwiftUI for all UI. SwiftData for local persistence — no server, no account required, full privacy. Claude API for the AI features (Focus Mode recommendations, insights generation). Xcode for build and distribution.

What I Learned

Designing a natural language parser that feels magical but behaves predictably is genuinely hard. The key insight was defining a strict grammar of intent patterns and mapping freeform input to that grammar, rather than trying to handle arbitrary text.