monospace.studio
← Back to Studio

Studio archive · 2025-06-19

When prompting goes wrong... and what we can learn.

A cautionary AI note about inaccurate prompting, messy outputs, and what those failures reveal about better product practice.

FAIL on using Cursor to setup N8N Workflows with Docker (Roughcuts)

In the spirit of oversharing and being human I paste here for all to see a poor UX designer’s (me) working struggles with AI prompting for building apps and automation. In this video my goal was to set up a basic JSON n8n workflow from Cursor - this fails… BUT we did get a ton of learnings on how far prompting for apps can go with an INSANE architecture setup… big learnings… so big win for me.

From failing to specify certain parameters the AI Agent went “rogue” but in a good way: explicit documentation and architecture for development; changelogs and shell scripts for extensive testing. Watch the video to see it live!

Onward!