Blog

One concern I had after our earlier work on collaborative vibe coding was that the task setup was still cleaner and simpler than many real coding interactions. Real-world vibe coding is messy: users differ a lot in how much they know, how much they trust the model, what kinds of failures they can actually notice, and how clearly they can describe those failures back to the agent.

That is the motivation behind vibe-coding-simulator, a project for turning vibe coding into a controlled multi-round experiment without stripping away the social structure that makes it interesting. The project repository is here: github.com/Haoyu-Hu/vibe-coding-simulator.

Everyday AI use comes with a familiar frustration: sometimes the model does not follow the original instruction, and repeated correction does not help. In the worst cases, the answer drifts farther away from the task, accumulates unnecessary changes, or becomes harder to trust turn after turn.

That failure mode is the motivation behind iterative-collapse-detection, a project for studying whether multi-round refinement actually improves model behavior or whether it can trigger a gradual collapse in quality instead. The codebase brings together a standardized multi-turn benchmark, reproduction runners, dataset preparation, evaluation, logging, and analysis utilities in one place. The project repository is here: github.com/Haoyu-Hu/iterative-collapse-detection.