LearnForge

Inspiration

Most beginners don't fail because they can't write code — they fail because they don't know what to build, what to learn, what to do next, why their code isn't working, or whether they're actually improving. Tutorials teach syntax; nobody teaches the journey. I built LearnForge to turn that confusion into a structured path: IDEA → PLAN → LEARN → BUILD → TEST → REFLECT → SHIP.

What it does

LearnForge is an AI-powered project-building mentor. Enter a software idea and it generates a full project plan, an interactive 6-milestone roadmap with 20+ tasks, level-aware Learning Mode lessons, a contextual AI mentor (hints, not solution-dumps), an educational "Debug With Me" coach, project health analytics, a learning journal with AI summaries, a visual Growth Timeline, XP/streaks/achievements, and a graduation report when you ship. A one-click Demo mode drops judges straight into a populated example project.

How I built it

Next.js 14 + TypeScript + Tailwind CSS + Framer Motion + Lucide on the frontend; Next.js API routes on the backend with all AI calls server-side so keys never touch the browser. Prisma schema (11 models) ready for PostgreSQL, with localStorage persistence for a zero-infra demo. The AI layer (lib/ai/) is provider-agnostic — Mistral and OpenAI behind one interface with a deterministic demo fallback, 8 dedicated system prompts, and structured JSON outputs (raw JSON is never shown to users).

Challenges

The biggest challenge was making AI output structured and reliable — solved with JSON-mode prompts plus a validation/normalization layer. Second was making the mentor teach instead of dumping answers — solved with hint-first, checklist-first prompting grounded in the user's actual milestone and history. Third was demo stability: the app had to survive judging with no network or credits, so every AI feature degrades gracefully to realistic demo responses.

What I learned

Structured AI outputs beat free-form text for product UX. Teaching prompts feel more intelligent than answer-bots. And demo design matters as much as features — the golden 3-minute path (idea → plan → roadmap → task → mentor → report) shaped the whole architecture.

Built With

• auth
• framer-motion
• markdown
• mistral-ai
• nextjs
• openai
• postgresql
• prisma
• react
• rest-api
• tailwind-css
• typescript
• vercel