Checkybear - AI Chess Coach

CheckyBear is an AI chess companion for kids that teaches chess through real-time voice, visual feedback, and personalized puzzles based on how you play.

Comment

Inspiration

Most chess apps feel overwhelming for kids and have confusing analysis. I wanted to build something that felt more like a friendly AI companion that teaches chess naturally while you play.

What it does

CheckyBear is an AI chess companion for kids.

It watches games in real time, explains mistakes, answers questions with voice, generates puzzles from your own gameplay, and can analyze real chess boards from a photo.

How we built it

We built most of the project using Medo.dev builder alongside a lot of planning and debugging with AI.

We used:

  • Large Language Model Plugin
  • Text-to-Speech Plugin
  • Speech-to-Text Plugin
  • The News plugin

We also generated many UI assets separately and uploaded them into Medo builder.

To save credits, we mainly used fast mode together with the prompt optimizer since it only cost one credit and helped improved output.

Challenges we ran into

AI consistency was one of the hardest parts. Sometimes fixing one issue caused another issue later, and the UI changed constantly during iteration.

We also had strange issues like random Chinese text appearing in outputs.

Adding:

“Only make the exact change described below.”

to prompts helped reduce unrelated AI changes a lot.

Another challenge was limited documentation for plugins, especially around TTS voices.

We also found that sending console errors to a seperate ai chat made debugging much easier. Direct console and network access inside medo.dev would honestly be a huge game changer.

Accomplishments that we're proud of

We’re proud that CheckyBear feels friendly and interactive instead of just another chess engine.

We managed to combine:

  • realtime coaching
  • conversational AI
  • voice interaction
  • adaptive puzzles
  • polished UI motion

into one hackathon project.

What we learned

We learned that building AI apps is mostly about iteration, prompting, debugging, and creating a good user experience.

We also learned how much voice, animation, and personality affect how intelligent an AI feels.

What's next for AI Chess Coach

Next, we want to massively improve speed and responsiveness since the app currently waits for AI output before passing it into TTS.

We also want:

  • login support
  • long-term memory
  • smarter personalization
  • synced voice + board highlights
  • more game modes
  • classroom support
  • school tablet integration

Long term, we want CheckyBear to become a free AI chess companion used by kids in schools everywhere.

Built With

  • medo
Share this project:

Updates