Skip to content

Latest commit

 

History

History
25 lines (21 loc) · 1.53 KB

File metadata and controls

25 lines (21 loc) · 1.53 KB

Phase 1: targeting teacher - student basic flow

  1. Teacher:

    • Upload videos to local SQL database
    • Annotate videos with milestone markers
    • Mock questions (with no AI feature yet)
  2. Student:

    • Watch interactive videos with milestone questions
    • Answer questions to progress through content
  3. Errors

    • Upon completed the video, the video preview reverts to old states with 0 correct answers and no question states
    • Need a plan to refactor video preview: at least the state is centralized, but the preview video still has a bunch of problems.

Problem 1: AI keeps overwriting working components

As AI keep overwriting working components, I need to understand which components it is intended to change based on .

  • The AI is very GRIT, it will slowly build the entire testing system to fix bugs. However, it needs architectural and coding principles guidance when it is too focused on one specific components and not the whole picture.
  • The AI is prone to tunnel-vision, requiring human aid in pointing out larger relevant context window
  • The AI is easy to mis-align with original plan and architecture, requiring human to remind them to stick with the plan.
  1. Confirm affected components? ask why?
  2. Asking iteratively until a clear picture is provided (may 2-3 times)
  3. Ensure alignment with reported architecture, need to constantly remind AI of alignment with initial plan
  4. Proceed with the plan
  5. After code change, demand second check to ensure that non-relevant logic is not changed