Updated demo video: expired API key and a few bug fixes
The first version of the demo video was recorded after the API key for the demo's LLM judge had expired. There were also a few bugs in the demo environment, including one that stopped the headless browser from launching. Because of this, the demo runs failed on screen while the narration described them succeeding. I renewed the key, fixed these bugs, and re-recorded the video against the live Cloud Run deployment. The new video shows both runs end to end: a clean submission that passes the sandbox tests and the LLM judge and gets its escrow released, and an adversarial submission whose embedded prompt injection is detected, so the escrow is rejected.
Log in or sign up for Devpost to join the conversation.