Actically

Inspiration

Getting an AI answer is easy. Understanding it well enough to explain it—and remembering it later—is harder. Actically connects these steps in one learning workspace.

What it does

Actically helps learners move from asking questions to actively practicing and reviewing:

• Guided learning: Explore a topic through Socratic questions, step-by-step solutions, or direct explanations.
• Learn from your materials: Organize study sets and paste reference sources.
• Feynman practice: Explain concepts in your own words and receive structured feedback.
• Blurting practice: Recall what you know before comparing it with selected references.
• Knowledge review: Review and approve extracted concepts before turning them into flashcards.
• Spaced repetition: Review due cards with FSRS scheduling and track progress using learning evidence.

The interface is designed for Vietnamese learners, with a consistent workspace across desktop and mobile layouts.

How we built it

Actically uses Next.js, React, and TypeScript for the application, Supabase for authentication and PostgreSQL storage, and Drizzle ORM for database access.

The server-side AI integration uses NVIDIA Nemotron through Nebius Token Factory. Structured output validation and source-reference checks help keep AI feedback grounded in the learning context.

Challenges and lessons

The main challenge was connecting chat, practice, knowledge extraction, and review into a coherent learning loop. We also addressed streaming cancellation, safe retries, private user data, and responsive interface consistency.

What's next

Complete live-service validation, test the full learning journey with learners, and improve feedback quality through educational evaluations.

Built With

• next.js
• react
• supabase
• typescript
• vercel