The best red pen, or a blunter knife?

#ensakidag 498: AI as the best red pen of all time. (Or just an increasingly blunt knife?)

Ethan Mollick took one of his very first research papers, almost exactly 20 years old, and ran it through GPT-5.5 Pro. Back came new data, razor-sharp tables showing what had happened since, and an analysis so sharp that someone in the thread called it the best red pen they’d ever seen.

That set my thoughts racing. Partly the concern in the comments: if we only let AI build on what already exists, the knife grows longer but blunter. And partly an idea I can’t let go of: what if we ran all of humanity’s accumulated knowledge through the models again, every time they took a substantial step forward? And what if you did it with your own life’s work, everything you’ve written and thought, and asked for honest feedback, or simply let it live on, continually updated?

Can it go wrong? Of course. But as always, it comes down to how we lead the work. Try it yourself and see.

And speaking of trying things yourself, on Wednesday you’ll see what I came up with after reading that tweet...

As always. Tell me what you think.

CC-BY 4.0 · and inspired by @emollick