Selected work · WeaveHacks · July 12–13, 2025
CodeTurtle
I built CodeTurtle with three teammates to catch the bugs that pass CI and still break in production. It reviews a pull request, opens the app, and tests it like a real user.
We built it at WeaveHacks in July 2025 and won a hackathon prize. My teammates were Ahmet Dedeler, Yelarys Yertaiuly, and Elliott Friedrich.
We later put it online at turtle.codes. That site is no longer running, but you can still see the demo and our code below.
Watch the demo
The team
From pull request to browser test
The idea was simple: code can look fine and pass its tests, while the actual app still has a broken login, a button that does nothing, or a confusing flow. We wanted to test what happens when someone actually uses it.
- Understand the change. Collect the pull request, code changes, and project documentation.
- Plan the checks. CrewAI agents generate focused QA scenarios and identify deployment information.
- Run the app. Browserbase and Stagehand automate browser interactions and capture screenshots.
- Review the results. A dashboard presents sessions and test results, while W&B Weave traces agent execution.
My part
I worked on the dashboard and the connection between the browser tests and the results you see. I built out the workflow view so you could follow each step, see whether it passed or failed, and open the screenshot for that step.
I also worked on screenshot capture and made sure test results stayed linked to the right repository. Each repository got a consistent ID, which connected the test runner, saved results, and dashboard sessions.
Some of my work on screenshots and the workflow view
Python · TypeScript · React · Next.js · CrewAI · Browserbase · Stagehand · W&B Weave · Supabase