Hi everyone,
I wanted to share a workflow auditing technique I stumbled upon recently.
We all know the struggle of trying to get a second pair of eyes on a complex n8n workflow. Sharing the JSON is great, but it requires the other person to import it and understand your logic. Screenshots are often too static.
I’ve been experimenting with Google NotebookLM as a code reviewer, and the results are surprisingly good if you use Video as the input.
The concept: Instead of pasting JSON, I record a 2-minute Loom/video of me hovering over the nodes and explaining the flow. I upload that to NotebookLM with a “Code Reviewer” system prompt.
Why it works for n8n: The AI processes the visual flow (the lines/connections) and the audio explanation simultaneously. It creates a “Red Team” audit that identifies logic gaps I missed.
Prompt and step by step instructions is available here : I Forced Google NotebookLM to "Roast" My Code. It Was Brutal. - DEV Community
I originally built this just to “roast” my bad coding habits, but it turned into a genuine debugging tool.
curious if anyone else is using Multimodal AI to document their workflows?