Prism Harbor Project | ai video made with ngram

About the video

This AI video was made with ngram to document a software engineering workflow captured directly from a developer's screen. The recording opens on the ngram web application showing its screen-recording and theme-selection interface, then shifts to the Cursor code editor inside a repository named studio-repo-3-ngram. There, the focus is a markdown file called fix-test.md that outlines an automated command for diagnosing and fixing failing end-to-end tests. The visible content walks through the command's input arguments, including TEST_FILE_PATH, TEST_CASE_NAME, and USER_CONTEXT, along with usage examples for running basic fixes, targeting a specific test case, and supplying context about recent code changes to improve fix accuracy. Critical configuration notes are shown, specifying the correct Playwright config file, the required test folder location, and the proper run command. A section on why user context matters is also visible, contrasting vague failure descriptions with specific, actionable ones. The final frame shows the same file open alongside a chat panel in agent mode using the Sonnet 4.5 model, with a list of past chat sessions related to reviewing code and deployment configurations. Overall, the recording serves as a walkthrough of an internal test-fixing workflow and its documentation.