AI did it in one prompt: Shopify CEO Tobi Lutke builds his own MRI viewer

Shopify chief executive Tobi Lutke has built a custom MRI viewer using artificial intelligence (AI) and his own medical scan data, showing how AI tools can be used to quickly create personalised software.

Lutke shared on X that he used Anthropic’s Claude AI to process DICOM files from a magnetic resonance imaging (MRI) scan. The AI organised radiology reports and images and generated a browser-based tool that ran locally on his computer.



The MRI viewer allowed him to navigate scans by body region, scroll and zoom through image stacks, and automatically highlight sections of the spine.

“This is a good example of reflexively reaching for AI,” Lutke wrote. “Once you work with these tools, it becomes natural to try them first. When I realised I needed Windows software while using a Mac, AI was the obvious solution.”

For context, DICOM is a standard file format used in healthcare to store and transmit medical images such as MRIs, CT scans, and X-rays. MRI uses magnetic fields and radio waves to produce detailed images of the body.

The post sparked a wider discussion online about how AI models are lowering technical barriers and allowing individuals to build tools that once required specialised expertise.

Netizens react

Several users noted that the viewer automatically marked findings directly on MRI images. Lutke responded that additional prompts enabled full annotation.

Others suggested improvements such as grid overlays and automated image review, which Lutke said he implemented immediately.


In related threads, some users shared similar experiments. One developer said he built a machine learning (ML) model to assess autoimmune thyroid risk using data from Apple Health.


Entrepreneur Bryan Johnson commented that better access to personal health tools could help individuals manage and improve their health.


Several users also asked how the AI was run from a USB drive. After being tagged, Grok explained that Claude was not running on the USB itself but was accessed via the cloud to process the MRI files and generate code, which then produced a local HTML viewer.