How a CS Graduate Built an AI Life Planner with Local LLMs and Switched to Linux Mid-Project

I Built an AI Life Planner the Month I Graduated and Switched to Linux Halfway Through

Hilal, a recent computer science graduate, built and deployed a full-stack AI life planner called Life Planner in the month they graduated. The app uses a locally hosted Qwen model via Ollama on an FRP home server for private, cost-free AI assistance and Google Gemini for lighter tasks.

Why This Matters

Most AI-powered planning tools rely on external APIs that incur costs and raise privacy concerns. By running a local LLM (Qwen 2.5:3b) on a home server and using Gemini’s free tier only for lightweight suggestions, this project demonstrates that private, low-cost AI integration is feasible for individual developers without powerful GPUs — challenging the assumption that production AI requires heavy cloud infrastructure or expensive subscriptions.

Key Insights

• Local LLM deployment: The app runs Qwen 2.5:3b via Ollama on an FRP home server, accessed over HTTP for fully private AI interactions at zero ongoing cost (2026).
• AI-native planning UX: Instead of manual form filling, users type natural language requests like ‘I want to study 2 hours every day’ — Aizen parses intent and creates actionable plan items with one click (2026).
• Claude Code terminal tool: Hilal used Claude Code directly in the terminal for coding, testing, and deployment — including automatic commit message generation — which accelerated development significantly compared to traditional workflows (2026).
• PWA as deployment strategy: The app includes PWA support via vite-plugin-pwa, enabling installable native-like experience from the browser without app store submission (2026).

Practical Applications

• Personal goal management: Users define life areas → yearly goals → monthly focuses → weekly priorities → daily tasks in a hierarchical planning system; pitfall: overcomplicating the hierarchy leads to user abandonment if onboarding is unclear.
• AI-assisted habit tracking: Natural language input via Aizen eliminates form friction for adding habits like ‘read 20 minutes daily’; pitfall: relying too heavily on AI parsing accuracy can produce incorrect entries if the model misinterprets ambiguous language.
• Multi-language support: Turkish language support was added after non-English-speaking test users couldn’t navigate the app; pitfall: assuming English-only UI is sufficient limits adoption among international users.

References:

• https://lifeplanner.hilalsay.com.tr
• https://github.com/hilalsay/LifePlanner
• https://medium.com/@hilalsaydogdu09/i-built-an-ai-life-planner-the-month-i-graduated-and-switched-to-linux-halfway-through-09f846ed8b56