AIYouTubeManager
Learn anything from YouTube, with AI — roadmaps, notes, quizzes and progress
An AI learning layer on top of YouTube: it builds a staged learning roadmap from a goal, matches videos to each stage, and writes notes, quizzes and flashcards from transcripts.
Status: Live with a 7-day free trial (Google sign-in), then Pro or Premium plans. Mobile app in development. · Period: Jun 2026 – present · Source: Private repository.
01Overview
YouTube is a huge free classroom without a syllabus. AIYouTubeManager adds the syllabus: type a goal and get a staged roadmap with weekly estimates, matched videos, AI notes, quizzes and flashcards, and progress tracking. It can optionally read your subscriptions and playlists. It does not download or host videos.
02The problem
Self-learners lose time to unstructured watching, forget what they watched, and cannot tell which of their subscriptions are actually worth their time.
03Who it is for
- Self-directed learners
- Professionals upskilling from YouTube content
- Students preparing for exams or new skills
04My role
Sole architect and developer: .NET back end, Next.js front end, AI integration, AWS infrastructure as code, and the CI/CD and rollback workflows.
05Implemented features
- Google OAuth sign-in and an optional YouTube connection that imports subscriptions and playlists.
- AI learning-roadmap generator with stages, weeks and estimated hours, plus videos matched to each stage.
- Per-video AI notes, key concepts, quizzes and flashcards written from the transcript.
- AI scoring of subscriptions and an AI-curated knowledge feed. Channels can be archived without unsubscribing.
- Semantic search and an AI mentor chat that uses RAG over the user’s channels.
- In-app embedded player with watch tracking.
- An RSS ingestion worker and a YouTube API quota-budget service.
06System architecture
A Next.js client talks to an ASP.NET Core API built on clean-architecture layers. PostgreSQL with pgvector supports retrieval, and Redis caches hot data.
07Engineering challenges and approaches
YouTube API quota limits
A background worker syncs new videos from YouTube RSS feeds, which cost no quota. A quota-budget service rations Data API calls so the product stays within YouTube’s limits.
Grounded AI answers
Videos and channels are embedded into pgvector. The mentor chat retrieves relevant content before generating an answer, instead of relying on the model’s memory.
Repeatable infrastructure
Terraform modules and bootstrap define the AWS environment. GitHub Actions run CI, Terraform plan/apply, environment deploys, diagnostics, Lighthouse checks and a rollback workflow.
08Outcomes and current status
- Live product with a free trial and paid plans.
- A clean-architecture .NET solution split into Domain, Application, Infrastructure and API projects, with tests.
No usage or revenue numbers are published here unless I can verify them.
09Roadmap (not yet shipped)
- Weekly digest email (Phase 2).
- Push notifications (Phase 2).
- Flutter mobile app (in development).