Note: This post is also shared on LinkedIn.

I recently recovered professional reports written over 40 years ago. Most pages were dense, hasty, curly scrawls—some written with a Montblanc 149—that I could barely decipher myself. In the past, these would have been “dead” files, trapped in ink. I photographed them and fed them into a modern LLM. The result? Clean ASCII transcriptions in seconds. We have come a long way since the LeCun-era MNIST days.

It made me realize something.

For many, handwriting remains one of the best front-ends for thinking and creativity. When you are truly stuck, you reach for a pen, not a keyboard; especially in mathematics. That has not changed in a century.

What has changed is the back-end. Today, longform handwritten thinking can flow directly into emails, documents, and more broadly into the digital world, without retyping or losing structure later. That works even for niche, jargon-heavy notes, because modern models rely on context, not perfect character recognition.

The toolchain evolved. The cognitive bottleneck did not. AI did not replace the pen. It made it interoperable.