VocabFlow

VocabFlow teaches IELTS vocabulary through real speaking tasks, not flashcards. Learn words with images, practice with AI conversations.

Comment

Inspiration

  • My wife works as a teacher who prepares students for IELTS exams and other English tests
  • Her work is rapidly changing because of AI advances and we wanted to implement a project that has long been in her dream of making language learning more entertaining and effective for students.
  • We're starting with using AI to help preparing for IELTS which is a more standardized yet an important test for students and professionals seeking to prove their English language skills.

What it does

  • Provides vocabulary learning through engaging use of images and immediate application using conversations
  • Task-based contextual learning - Students learn vocabulary through real IELTS speaking tasks like "Describe a person you admire" rather than isolated flashcards
  • Visual-contextual associations - Each vocabulary word is presented with contextual images to create stronger memory connections
  • AI-powered conversation practice - Using Eleven Labs, students practice vocabulary through natural conversations that simulate IELTS speaking scenarios
  • Intelligent spaced repetition - Smart review system that maintains task context while optimizing retention intervals for 80% vocabulary retention
  • Adaptive band progression - Students progress through IELTS Band levels (6-9) with placement testing and 80% completion requirements
  • Comprehensive analytics - Detailed progress tracking showing vocabulary mastery, learning velocity, and areas needing improvement
  • Mobile-first design - Clean, professional interface optimized for daily learning sessions

How we built it

  • Used Claude to be a prompt manager and tried to splice the prompts into seven key prompts using best practices gathered through Perplexity research
  • Used Bolt.new to build the project - used discussion mode extensively to let Bolt give a plan before implementing
  • Used Claude and Perplexity to help Bolt get out of its loop when solving bugs (we encountered a lot)
  • Strategic prompt engineering - Created systematic prompts that built upon each other: Foundation → Visual Learning → Progress System → AI Integration → Advanced Features → Design Polish
  • Content development - Generated sample IELTS tasks and band-appropriate vocabulary words with bilingual support (English/Russian)
  • Modern tech stack - React with Tailwind CSS for responsive UI, Supabase for data management, Eleven Labs for premium AI speech synthesis

Challenges we ran into

  • Various Supabase errors and bugs around data loading and real-time updates
  • Problems initially with using Eleven Labs since it defaulted to the system text-to-speech capabilities
  • Complex state management - Managing vocabulary progress, spaced repetition schedules, and conversation states across multiple components
  • Audio integration - Synchronizing speech recognition, AI responses, and vocabulary tracking in real-time conversations
  • Content organization - Structuring vocabulary words across themes, bands, and difficulty levels while maintaining learning progression

Accomplishments that we're proud of

  • Making the integrations work, especially using Supabase to load all the learning materials
  • Using prompts based on advanced prompting techniques and best practices
  • Having something tangible that can support us in our respective professions (my wife being an IELTS teacher and me being an amateur vibe coder)
  • Achieving the core vision - Successfully implementing task-based vocabulary learning with visual associations and AI conversation practice
  • Premium user experience - Creating a "Notion meets Slack" design aesthetic that feels professional yet approachable
  • Comprehensive learning system - Building a complete platform with placement testing, progress tracking, and adaptive learning algorithms
  • Real educational impact - Addressing genuine problems in IELTS preparation with innovative solutions

What we learned

  • Planning the prompts and inspecting them beforehand
  • Having good documentation and ensuring that Bolt follows the documentation and repeats it
  • Systematic development approach - Breaking complex features into manageable, sequential prompts prevents scope creep and maintains quality
  • AI tool orchestration - Effectively combining Claude for planning, Perplexity for research, and Bolt for implementation creates powerful development workflows
  • User-centered design - Starting with real teacher and student pain points leads to more meaningful feature prioritization
  • Content is king - High-quality, contextual learning materials are as important as technical implementation

What's next for VocabFlow

  • Product research - Test out with actual students to gather feedback and iterate
  • Research partnerships - Collaborate with language learning institutions to validate effectiveness and gather learning outcome data
  • Multi-theme expansion - Add Places, Experiences, and Objects themes to complete the full IELTS vocabulary spectrum
  • Teacher dashboard - Create tools for IELTS instructors to track student progress and assign targeted practice
  • Community features - Add peer learning, progress sharing, and study groups for collaborative learning
  • Mobile app development - Build native iOS/Android apps with offline capabilities and push notifications
  • Marketplace integration - Connect students with certified IELTS tutors and preparation courses

Built With

Share this project:

Updates