Vibe Coding in the Wild: How Shopify CEO Tobi Lütke Used AI to Solve a Real Problem
Shopify CEO Tobi Lütke shared a simple but powerful example of vibe coding: instead of installing Windows to open MRI data, he used AI to generate a clean HTML viewer—and then iterated it to add annotations. The real lesson isn’t the tool, but the reflex to reach for AI first and experiment without overthinking.
Vibe Coding in the Wild: How Shopify CEO Tobi Lütke Used AI to Solve a Real Problem
Sometimes the best examples of AI aren’t flashy demos—they’re small moments where AI removes friction in a way that feels obvious in hindsight.
Shopify CEO Tobi Lütke shared one of those moments: he receives his annual MRI scan on a USB stick, but the data can only be opened using commercial Windows-only software.
He’s on a Mac.
The “reflex” that changed the outcome
Instead of installing Windows (or searching for alternative tools), Tobi did the simplest thing he could think of:
- ran Claude on the USB stick
- asked it to generate a basic HTML-based MRI viewer
It worked.
And the result looked cleaner than the official software.
Then he pushed it one step further:
One more prompt—and it annotates everything with the findings.
No long planning cycle. No scope. Just quick iteration.
What vibe coding really means
The interesting part here isn’t the MRI viewer.
It’s the mindset.
Tobi described this as reflexively reaching for AI: after you tinker with AI enough, you stop treating it as a “special tool” and start treating it like a default option.
You don’t ask:
- “Is there an app for this?”
- “Should I make a project out of it?”
You ask:
- “What if I just try prompting it?”
That’s vibe coding at its best: lowering the activation energy between a problem and a working solution.
Why this matters for builders (and teams)
This kind of reflex changes how you operate:
- Less waiting for the “right tool”
- Less over-planning
- Faster feedback loops
- More experiments that actually ship
AI becomes a bridge between intent and execution: a quick prototype partner that helps you move from idea → working artifact in minutes.
How to train the “AI-first” instinct
You don’t train this on big, high-stakes projects.
You train it on small annoyances:
- “This file format is annoying.”
- “This workflow is slow.”
- “This tool only runs on Windows.”
- “I just need a simple viewer/exporter/converter.”
Low stakes. Fast iteration. Repetition.
Over time, the reflex becomes automatic.
Closing thought
The takeaway isn’t “AI should do medical things.”
The takeaway is simpler:
If you build with AI regularly, you start seeing friction as optional.
That instinct is valuable—and it’s a muscle you can train.