Skip to lesson
Exit
Writing, Demos & Developer Distribution1 / 3

2 min lesson

A concrete demo script that travels

Describe what "The whoa is the board-column trigger" changes in practice.

Step 1 of 3

A concrete demo script that travelsreal beats, not abstractions

Here is a worked example you can adapt, built from Cursor's own field demos so the beats are real rather than invented. The whoa is the board-column trigger: in Jira, add a dedicated column - call it In Progress (Cursor), split from In Progress (Human) so agent work is visible - wired via webhook so that dragging a ticket into it spins up a cloud agent that extracts the ticket, implements it and returns a PR. The single drag collapses what was a ~7-step manual ticket workflow.

Learn more

Full explanation

Proof-of-work is what made async demos credible

  1. 1Open on the drag. First 15 seconds: drag one ticket into the In Progress (Cursor) column. A cloud agent boots in an isolated VM. That's the beat - "wait, the board is the trigger?"
  2. 2Show the fan-out. Now drag four low-hanging-fruit tickets in at once. Four cloud agents spin up in parallel (LB1, LB2, LB3, LB4), each working its own ticket and opening its own PR. Parallelism is the part that makes a viewer lean in.
  3. 3Land on the proof, not a claim. Each agent comes back with an artifact: the PR diff plus screenshots and a screen recording of it testing the change in a browser. Show the video uploaded straight onto the PR. The proof-of-work is what makes "it shipped a feature" believable instead of magic.
  4. 4Narrate one judgment call. Point at where the plan surfaced a discrepancy - e.g. the agent paused because the ticket assumed authenticated users but it found no auth in the app, and asked how to handle the gap. That moment teaches that planning catches problems before code.
  5. 5Tie to outcome. End on the developer payoff: a backlog of one-shottable tickets clears itself while you review diffs, and human vs. agent work stays legible because they live in separate columns.
Proof-of-work is what made async demos credible

The reason this demo lands where a screen recording of local editing wouldn't: artifacts. A cloud agent navigates its own virtual desktop, makes a change, spins up a browser, tests it, records a video and screenshots, and self-corrects on console errors - then returns the walk-through video plus type-checks and test results, and can attach the video to the PR. Releasing artifacts caused a real uptick in adoption precisely because developers will trust async work they can watch verify itself. Your demo should foreground that proof, not the GIF of the agent typing.