Freelance designers lose billable hours not to complex projects, but to the subtle accumulation of 'one small extra screens' and casual deadline shifts. Manage Your Tasks, a new open-source submission for the Hacktoberfest Open-Source AI Challenge, tackles this scope creep by running Google’s Gemma 3 1B model locally on the user’s machine. Built by TonySama-BoiledCabbage and DD998855, the tool parses client feedback from WhatsApp, email, and calls to flag potential requests and dates, ensuring every AI suggestion is pinned to the original source text for human approval.

The Local-First Architecture

The core technical differentiator is the elimination of cloud inference costs. By utilizing Ollama to run the open-weight Gemma 3 1B model, the developers have removed the per-token billing that forces most AI tools into subscription models. The application stores project metadata in browser localStorage and original files in IndexedDB, with parsing and inference handled by local HTTP endpoints. If Ollama is not running, the app explicitly states this rather than silently failing, and the rule-based detection mode is clearly labeled as a non-AI fallback.

Business Model and Privacy

This zero-marginal-cost inference structure enables a 'pay once, own it' business model that directly addresses freelancer subscription fatigue. The free tier includes projects, tasks, and rule-based detection, while the Pro tier—sold as a one-time purchase—unlocks the local Gemma semantic understanding for indirect asks and mixed English/Chinese text, along with scope-change approval records. The team explicitly states that privacy is the product; since client feedback often contains unreleased designs and NDA-bound pricing, keeping data on the designer’s laptop mitigates the risk of sending sensitive material to hosted APIs.

Validation and Limitations

The development team is transparent about the current state of the project, admitting that they have not yet validated pricing or landed a paying customer. The AI’s role is deliberately narrow: Gemma classifies segments as possible requests, but a deterministic resolver must verify that extracted date phrases exist in the source text before converting them to calendar dates. Ambiguous inputs are left blank for manual entry. Known limitations include potential misses on indirect requests and unreliable drag-and-drop task creation on the timeline, with no current metrics on AI precision or recall against a representative dataset.

Key Takeaways

  • Manage Your Tasks uses Gemma 3 1B via Ollama for local, offline inference, eliminating cloud API costs.
  • The tool enforces a human-in-the-loop workflow where AI suggests tasks but users must explicitly approve them.
  • A one-time Pro purchase model is viable specifically because the open-weight architecture removes recurring inference expenses.
  • The project is a Hacktoberfest submission and has not yet undergone commercial pricing validation or large-scale user testing.

The Bottom Line

For builders and freelancers drowning in subscription tools, Manage Your Tasks offers a refreshing, privacy-first alternative that proves local AI can power sustainable business models without the cloud tax.