-
-
Workspace: Your code on the left, your understanding on the right. Explore Goldie’s IDE-shaped web walkthrough.
-
Sample code: Follow a prepared team-management build and explore its frontend, API, database schema, and architectural handoff.
-
1. Practice: Explain your reasoning and compare it with a model answer. Practice UI shown with example questions and answers.
-
Lesson: Understand the decisions behind the code, one concept at a time. Lesson UI shown with example teaching content.
Inspiration
We’re vibe coders who want to become more technical. We’d give an AI agent a prompt, then sit there scrolling while we waited for it to finish. When the code arrived, we had a working feature but often couldn’t explain how it worked. Goldie came from a simple idea: what if that waiting time became learning time?
What it does
Goldie is an opt-in engineering mentor inside a desktop IDE. It teaches concepts like frontend state, APIs, and databases through short spoken lessons. Users can pause, skip what they know, ask a question by typing, and return to their lesson. They can keep coding and learning at their own pace.
How we built it
We forked the open-source Eclipse Theia IDE and built Goldie’s lesson experience into it. We created a native lesson panel and a backend that manages lessons, questions, progress, and audio. ElevenLabs generates the spoken explanations. We also built a Gemini-powered backend that can create lessons from a supplied project summary.
Challenges we ran into
One challenge was designing lessons for very different prompts. “Build me an app” calls for a broad architectural overview, while a small, specific request may need just one focused concept. We also wanted users to complete their coding task and their lesson independently, without either one interrupting the other.
Accomplishments that we’re proud of
We built a working spoken lesson experience inside a real IDE. Users can hear audio generated through ElevenLabs, pause and resume it, skip a phase, ask a typed question and hear the answer, then finish with a recap. Bringing all those pieces together made Goldie feel like a product we’d actually use.
What we learned
We learned how to integrate the ElevenLabs API into an interactive app and manage audio alongside lesson progress. We also learned that teaching AI-generated code means explaining the decisions behind it, not just describing what each line does.
What’s next for Goldie
Next, we want to connect lessons directly to live coding activity and bring Goldie into tools people already use, including Codex, Cursor, and Claude. Our goal is to let people keep building with their favorite agent while gaining the technical understanding to take ownership of what they create.
Built With
- codex
- eclipse-theia
- electron
- elevenlabs-api
- fastify
- gemini-api
- node.js
- react
- typescript
Log in or sign up for Devpost to join the conversation.