A new open-source-style project called Didon is aiming to solve the eternal problem of time tracking without the manual overhead that makes most tools unbearable to use. The macOS app runs quietly in the background, capturing periodic screenshots and using a local LLM to analyze exactly how you're spending your workday โ€” generating structured work journals automatically with zero cloud dependency.

How It Works

Didon operates in three stages: First, users define their project goals and activity categories as context for the AI. Second, the app monitors screen activity through periodic screenshots while running unobtrusively in the system tray. Third, a local instance of Qwen-3-VL:2b processes the captured data on-device, extracting task details like window titles, file names, active applications, and detected work items โ€” then maps everything to your predefined project structure.

Privacy-First Architecture

What separates Didon from enterprise surveillance tools is its commitment to keeping sensitive data local. The app explicitly states that screenshots and activity logs never leave your Mac. All AI analysis happens via the Qwen-3-VL:2b model running locally, which means there's no subscription cost for API usage and no risk of corporate data leakage. It also respects system sleep โ€” tracking pauses automatically when your machine idles so you don't end up with phantom work hours logged at 3 AM.

Features and Pricing

The feature set includes automatic daily log generation with CSV export for billing or reporting purposes, smart context-aware categorization that ties activity to your specific projects, weekly trend analysis showing peak productivity hours and time drains, and system startup integration so tracking begins immediately. Early access pricing sits at โ‚ฌ89 (reduced from a crossed-out โ‚ฌ199) for lifetime access and updates with one-on-one support included.

The Indie Developer Problem

The creator's motivation resonates with anyone who's tried to optimize their solo workflow: traditional time trackers demand constant manual input, physical timers feel clunky, and neither gives you meaningful feedback about where hours actually disappear. Didon emerged from that frustration โ€” the realization that 'marketing, consuming content and unnecessary activities kept stealing hours' without any way to see the pattern until end of week.

Key Takeaways

  • Uses Qwen-3-VL:2b for on-device AI analysis with no cloud processing or data transmission
  • Generates structured work journals automatically from screen monitoring
  • CSV export available for client billing, internal reporting, and productivity audits
  • Early access pricing at โ‚ฌ89 lifetime with Windows version reportedly in development

The Bottom Line

Didon represents the kind of tool that only becomes possible when local AI models get small enough to run everywhere. For indie developers and privacy-conscious teams who want quantified work habits without surrendering their screen activity to third-party servers, this approach hits the right balance โ€” though periodic screenshot capture will give some users pause about adoption in shared or corporate environments.