ToTheCore

You want the answer but you still want more. This app is for the hungry learners who aren't satisfied with just the answer.

Comment

Inspiration

I love learning, but learning can be frustrating. There is nothing worse than struggling with a concept and nothing better than finally understanding it.

I noticed something while learning: YouTube videos that use analogies are much easier to understand. Difficult concepts in everyday life become easier when you break them down to the basics and relate them to things we encounter every day.

This inspired ToTheCore, which captures the fundamental characteristics of effective learning:

  • Breaking things down to their root concepts to avoid confusion.
  • Using simple, everyday analogies to make abstract ideas concrete.

The frustrating part of learning can now be bypassed—students and learners can finally understand the core of any concept.

What it does

ToTheCore accepts any type of input:

  • Typed text
  • Uploaded files (PDF, DOCX, images)
  • Voice recordings

It then:

  1. Normalizes the input into text.
  2. Breaks concepts down to first principles.
  3. Explains step-by-step using child-level analogies.
  4. Highlights common points of confusion.
  5. Teaches before showing formal answers, including math:

For example, if a user submits a physics problem with the formula:

[ F = m \cdot a ]

ToTheCore would explain:

  • What force, mass, and acceleration mean.
  • An analogy: “Pushing a shopping cart: the heavier the cart ((m)), the harder you need to push ((F)) to move it ((a)).”
  • Step-by-step reasoning before showing the calculation.

How I built it

Frontend:

  • React + Tailwind for a clean, distraction-free UI
  • Components: InputBox (for text), StepViewer (step-by-step display), ModeSelector (choose simplicity level)

Backend:

  • Node.js + Express server
  • Routes: /explain for text, /voice for audio uploads
  • Speech-to-text: Whisper API converts voice to text
  • File parsing: PDFs, DOCX, and images converted to text
  • Gemini 3 integration: All input is wrapped in a teaching prompt and sent to Gemini for stepwise explanation

Key architecture concept:

  • Everything, no matter the format, becomes: Analogy: “All confusion is funneled into the AI, which rebuilds understanding from the ground up.”

Challenges I ran into

  • Multi-input handling: Managing text, files, and voice in one pipeline was tricky.
  • File parsing: PDFs and images required OCR integration.
  • Prompt design: Crafting the Gemini prompt to teach rather than just answer.
  • User experience: Showing explanations step-by-step without overwhelming the user.

Accomplishments that I am proud of

  • Successfully built a multi-input learning assistant capable of teaching anything.
  • Created a prompt framework that consistently produces analogies and stepwise explanations.
  • Implemented a voice-to-text pipeline using Whisper for real-time spoken input.
  • Made the system child-level understandable, bridging the gap between answers and comprehension.

What I learned

  • Teaching AI to teach, not just answer, is different from building a standard chatbot.
  • Analogies are powerful: simple, everyday references drastically improve comprehension.
  • Multi-format input systems are feasible for a hackathon project but require careful normalization.
  • Explaining math or code step-by-step can be done incrementally, ensuring clarity without oversimplifying.

What's next for ToTheCore

The next steps will be:

  1. Make ToTheCore cross-platform, available on both desktop and mobile.
  2. Add interactive explanations, where users can ask “go deeper” or “switch analogy” dynamically.
  3. Integrate real-time collaborative learning, so multiple users can explore a concept together.
  4. Optimize for speed and responsiveness, especially with large files or long voice recordings.

Built With

Share this project:

Updates