It's been about ten days since I wrapped up an in-house seminar. Titled 'The Printing Revolution of the AI Era', it used the printing revolution of the 1450s as a mirror for looking at the AI revolution we're living through now. The talk itself went fine, but what stayed with me wasn't the slides — it was a single thought that quietly settled in while I was preparing them. Perhaps this post is my attempt to slowly unpack that thought.
It started with a video. Boris Cherny, the engineer who built Claude Code, said something in a talk: if you had to pick the one moment in the history of technology that most resembles what's happening now, it would be the printing press of 1400s Europe. It was the kind of remark you could easily let slide, but strangely, it wouldn't leave my head. I happened to be preparing the seminar anyway, so I figured that while I was at it, I'd dig into this analogy all the way down. Did the printing revolution of the 1450s really resemble today's AI revolution? And if so, where did that revolution end up?
Before the printing press, a book was something copied out by hand, one volume at a time. Producing a single Bible reportedly took parchment from 200 sheep, quills from dozens of geese, and eighteen months of a scribe's labor. Books were precious, naturally, and writing stayed in the hands of a few — clergy, scholars, nobles. Then Gutenberg's press arrived in the 1450s, and the landscape changed. The number of European cities with a print shop went from zero to 236 in fifty years, book prices fell by nearly 75 percent, and the number of books coming into the world exploded — exponentially, in the literal sense. All the press did was take the scribe's seat between the author and publication, yet the total amount of writing in the world swelled to an entirely different scale.
What if we lay this scene right over the 2020s? People who can read and write code are still a minority. Supposedly only around 0.5 percent of the world can code — a figure that oddly echoes the statistic that about 10 percent of adult men in England around 1500 could sign their own name. So for a long time, programs were scarce, expensive things, stitched together one at a time by a small number of trained hands. But now, AI has begun moving into the programmer's seat between the author and distribution. Designers, accountants, operations people — folks who barely know code — are building their own workflows and using them. This was the biggest insight I gained while preparing the seminar. AI is the printing press of the digital age. Just as the press printed text, AI prints code — and if that's so, might the number of programs explode exponentially, the way books once did? The very stretch where book supply exploded in the fifty years after the press — we may be passing through it right now, just two or three years after ChatGPT made generative AI mainstream.
So will programmers vanish the way scribes did? The deeper I dug into the history of print shops, the more I saw the opposite scene. The early printer was a one-man startup: he cast his own type, set the pages, inked them, ran the press, proofread, bound the books, and sold them himself. But as books started pouring out, the work inside the print shop gradually split apart — and in that process, interestingly, a few professions, far from disappearing, actually rose to prominence. Printing was an enormously expensive business, so the publisher emerged to decide which books to print. And because the same error would be copied unchanged into thousands of copies, the editor emerged to select and refine manuscripts, and the corrector to check proof sheets against the original. In 1472, one print shop in Rome misread demand so badly that it was left with 12,475 unsold books and even sent a petition to the Pope. The easier copying itself became, the more precious grew the eye that judged what to print and the eye that verified it was printed right.
I've come to think this scene shows us, in advance, where engineers will stand in the AI era. What vibe coding lowered is the barrier to entry for programming. Anyone can print out code now, but for that code to survive as a real service, it has to meet technical bars like speed, stability, and security — and above all, it has to be something someone actually wants. Just as the editor and corrector did in the print shop, might judging what to build and verifying it was built right emerge as the engineer's new expertise in an era when programs pour out? After the seminar, I came away with a conviction of my own: in the AI era, the engineer's role will matter more, not less.
By the time the printing revolution wound down, they say, the world had become a place where everyone could read and write. So by the time the AI revolution winds down, will we be living in a society where anyone builds and uses programs of their own? What kind of engineer will I have become by then, and what will our child be building to make a life in that world? I find myself quietly setting the two futures side by side and imagining them.