Description:Watch current coding agents tackle the same ridiculous build prompt and see what unfolds. While they work, we throw curveballs, compare results, and discuss the harness that keeps the fun contained: disposable workspaces, cloud and local agents, tests, reset paths, and human review. The winner is judged by usefulness, chaos recovery, laughs, and whether we'd dare run it again.
What you will learn:- Run coding agents in disposable workspaces so experiments do not touch your everyday computer or production repo.
- Design harmless, funny build prompts that reveal code quality, test behavior, tool use, and recovery from changing requirements.
- Score builds based on their ability to run, recover from curveballs, stay within safety harness, and make the room laugh.
- Compare cloud-hosted and local agent trade-offs for setup speed, tool access, privacy boundaries, cost, and reset behavior.
