Babbleborg Build Guide - About This Guide

What Makes This Approach Different

Instead of an off-the-shelf product with per-student monthly fees and privacy concerns, this guide teaches you to build your own version using AI coding assistants. This means:

✅ You understand the architecture deeply (not just copy/paste)
✅ You can customize it for your specific needs
✅ You can support and maintain your own code
✅ You work with current AI tools, not old code
✅ You gain valuable experience working with AI coding agents

Philosophy

This guide recognizes that AI coding assistants are now powerful enough to help technically proficient non-programmers experiment with building useful, working applications.

You don’t need to be a professional developer. You need to:

Understand what you want to build
Think architecturally about how systems fit together
Test methodically and read error messages
Iterate with your AI assistant when things don’t work

This guide provides the architectural roadmap. Your AI assistant provides the implementation support. Once you’ve done this, you’ll have a proven process for tackling other projects, too.

This is a Guide

This site is a guide, not a supported product. You’re building your own implementation, which means you have:

✅ Complete ownership and control over your code
✅ Freedom to modify and extend as needed
✅ An understanding of how everything works

It also means that you’re largely on your own, which is not a good fit for every person or school. There are plenty of high-quality, well-supported AI products out there if you’re not in an environment where experimentation and iteration are acceptable. They’re just not as cool as making your own.

API Training Privacy

API-based services (Anthropic, OpenAI, Google Gemini paid tiers) do not use customer data for model training as of October 2025. Consumer chat products have different policies. This distinction is why the guide emphasizes API-based deployment over consumer chat interfaces. Always verify current terms with your chosen provider.

Real-World Costs

A real classroom conversation with an eighth-grade student planning a technology project (34,592 input tokens, 1,893 output tokens) cost $0.0052 using Gemini 2.5 Flash - about half a cent. Even premium models like Claude 3.7 Sonnet cost only $0.11 per conversation.

As of October 2025, a student having 10 conversations per month would cost $0.05 to $1.10 depending on the model chosen - far below commercial AI tools that charge $5-15 per student per month regardless of actual usage. See current pricing at llm-prices.com.

License

The written content of this guide (including text, images, diagrams, and downloadable context/prompt files) is licensed under Creative Commons Attribution 4.0 International (CC BY 4.0). You're free to share and adapt — even commercially — with attribution. If you make changes, please indicate that changes were made.

How to Attribute:
"Adapted from Babbleborg Build Guide by Tom Gromak (CC BY 4.0)" with a link back to https://babbleborg.org when possible.

Disclaimer:
This guide is provided "as is," without warranty of any kind, express or implied. You are solely responsible for any applications or outcomes that result from using it.

Acknowledgments

This guide and the original Babbleborg were built with extensive assistance from Claude and inspired by the work of Ethan Mollick and Simon Willison. This project demonstrates what’s possible when educators combine pedagogical expertise with modern AI development tools.