Inspiration
Modern development is shifting toward “vibecoding” building quickly with multiple AI tools like Gemini, ChatGPT, and Claude. But as speed increases, clarity drops. After a few prompts and edits, we kept asking ourselves: “What did we just change and why?” Even worse, AI was writing code faster than we could understand it. We wanted something that doesn’t just track changes but actively explains what’s happening as you code to your preference style.
What it does
DevLog is an AI-powered development memory layer that tracks, explains, and narrates your code in real time. As you code, DevLog:
- Watches file changes instantly
- Uses Gemini to understand and classify diffs
- Generates plain-English explanations of what changed and why
- Logs everything into a living DevLog stored in Firestore
- Visually updates a dashboard and VSCode extension in real time
- Optionally provides auditory feedback (via ElevenLabs) to tell you what just happened -DevLog already acts as a unified memory layer across multiple AI platforms (Gemini, Claude, ChatGPT, Cursor), keeping all decisions and changes in one place.
Instead of guessing, you can literally see and hear your code evolving.
You can also:
- Ask questions like “What broke last?”
- Track decisions across AI tools
- Generate instant handoff documents
How we built it
We built DevLog as a real-time, multimodal system using Google Cloud + Gemini:
- Gemini → diff understanding, summarization, classification
- FastAPI backend (Cloud Run) → handles change processing + APIs
- Firestore (Firebase) → real-time sync across all clients
- Watchdog file watcher → detects local code changes
- React dashboard → visual timeline + DevLog
- Code extension → in-editor insights and status
- MCP server → connects external AI tools like Claude Code
- ElevenLabs → converts AI explanations into live audio feedback
Challenges we ran into
None of us were super experienced with using Google Cloud APIs going into this project. We had to quickly learn how to work with Gemini AI, Firestore, and Cloud Run under tight time constraints. Understanding how these services connect especially around authentication, deployment, and real-time data syncing was a challenge at first. But by iterating quickly and learning on the fly, we were able to build a fully integrated system that leverages multiple Google Cloud services together.
Accomplishments that we're proud of
Built a real-time, multimodal developer feedback system Code changes are not just tracked but also they’re explained visually and audibly Fully integrated Gemini + Firebase + Cloud Run End-to-end system working live across backend, extension, and dashboard A demo that feels alive where you can see and hear your code evolve
What we learned
Planning and clearly assigning tasks made a huge difference. Early on, we defined clear ownership across backend, extension, and frontend, which allowed everyone to move in parallel without stepping on each other’s work. Having a shared structure, API contracts, and agreed-upon schemas kept everything aligned and prevented last-minute integration issues. It turned what could have been chaotic into a coordinated build.
What's next for DevLog
- Shared DevLogs across teams so everyone sees the same project memory
- Predict bugs before they happen based on patterns
- Long-term memory using Gemini to understand patterns across sessions
- Auto-detect architectural issues and suggest improvements -Tag teammates in decisions and track ownership
- Learn your workflow and adapt over time
Log in or sign up for Devpost to join the conversation.