Hatch Coding

About
the Client

Hatch Coding is an AI-powered learning system that teaches students as young as eight how to code with no prior experience necessary. Students invest 2-3 hours per week across four certificates to develop the skills needed to pass a technical Jr. Developer interview. Hatch Coding has developed courses in JavaScript, Python Team Coding, Unity Game Engine, and Web Development to make their students proficient developers.

“Very responsive on communication, paid attention to budget and timeline. All was good for this project.”

Peter Kuperman CEO & Founder 
CEO & Founder

The
challenge

Hatch Coding sought a strategic partner to position their product competitively, improve the current platform, increase the conversation rate, and introduce new AI features. The goal was to teach children and teenagers how to code by leveraging AI, which would accelerate their learning and help students gain insights, hints, and solutions during challenging coursework. To address these goals, Rootstrap proposed reworking the current platform through the lens of AI to position their product.

We developed two main features to enhance the AI capabilities of the platform. While Integrating OpenAI within a Python project was straightforward, we encountered several AI-related challenges that were addressed.

What we did?

‘Solve with AI’ Functionality

Enables students to request AI assistance for code generation.
Provides a glimpse into AI's problem-solving and code-generation capabilities.
Demonstrates multiple problem-solving approaches.
As AI is not infallible, students learn to identify and correct potential errors or imperfections in the generated code.

‘Solve with AI’ Challenges

Generate code compatible with the Hatch coding IDEsome text.
- Used specific syntax rules and the P5 library.
- Translated these rules into natural language for GPT to follow.
- Ensured clarity, conciseness, and emphasis on critical pointssome text.
  - AI would highlight key points in the generated code that can be improved/changed.
- Simplified the functionality so it’s accessible for all ages.

‘Explain with AI’ Functionality

Allows students to select code segments, triggering an option to consult the "Explain with AI" assistant.
Analyze and provide clear, educational explanations for each code section
- A section could range from a single line to an entire function.
Enhances understanding of code syntax, structure, and semantics for users

‘Explain with AI’ Challenges

Initially, we instructed the AI to separate explanations with a specific character.
- At times, it failed to adhere to this structure, so we utilized OpenAI's new JSON response feature to resolve this, ensuring structured responses.

Built with Security in Mind

The user-input aspect of "Explain with AI” involved a risk of prompt injection.
- To mitigate this, we employed various prompt techniques, like few-shot examples, to guide GPT's responses in specific scenarios.
With Hatch coding's young audience, unrestricted GPT responses were not an option.
- We leveraged OpenAI's Moderation feature to detect inappropriate content or misuse, ensuring safe and appropriate usage.
We also addressed errors associated with the use of the OpenAI API.
- We tackled possible error responses, such as an unavailable API or inappropriate response, and displayed user-friendly messages.

Integrating with GPT

We performed market research and introduced GPT-4 Turbo to the application.
- Compared to GPT-4, GPT-4 Turbo had faster response times while maintaining the output quality.
- Availability issues with GPT-4 Turbo made it unsuitable for our production deployment timeline, but it was a promising candidate for future project iterations.
Enhanced JSON response format
- Successfully implemented the "Explain with AI" feature migration to the latest version of the OpenAI.

Hatch Coding

Services Provided

The Team

About
the Client

“Very responsive on communication, paid attention to budget and timeline. All was good for this project.”

The
challenge

We developed two main features to enhance the AI capabilities of the platform. While Integrating OpenAI within a Python project was straightforward, we encountered several AI-related challenges that were addressed.