Code / Choice grew out of a collaborative study of code-generation models. It makes a concrete question easy to explore: given two implementations of the same task, which would you rather maintain?
Public Play presents a pair, records a choice of A, Equal, or B, then reveals the sources. A separate, access-controlled Study supports the research workflow.
My contribution
- Built the React comparison interface, including readable code panels and mobile layouts.
- Implemented calibration, joint review, and independent review, with fixed assignments once formal review begins.
- Added a Node.js and SQLite service for saved progress, plus offline reviewer-file transfer and validated imports.
- Implemented result exports and kept exploratory Play responses separate from formal research data.
Scope
This is the application supporting the research, not a second independent research study. Public preferences describe what visitors choose when inspecting code; they do not establish correctness or general code quality.
The live app and source are public. Access to the formal Study and its response data is separate.