In the 12th century, cathedral builders in Europe set shards of colored glass into lead frames. Their goal wasn’t climate control. It wasn’t productivity. It was beauty and worship: a kaleidoscope of saints, angels, and cosmic geometry to make heaven feel close.

And yet—those windows did more than dazzle. Their uneven textures diffused glare, calming the eye. The metal oxides that made those deep blues and reds also happened to filter portions of the infrared spectrum, cooling interiors a touch. It wasn’t intentional. It wasn’t even noticed at the time. But centuries later, scientists studying those same panes realized the art carried function.

Art preceded utility. Beauty came first. And only much later did we name its practical effects.

Twain’s Passport and Einstein’s Encirclement

Mark Twain once wrote that “Travel is fatal to prejudice, bigotry, and narrow-mindedness.” His point wasn’t that passports should be issued as anti-bias programs. Travel was for trade, for adventure, for curiosity. The side effect—the shattering of narrow views—was the greater gift.

Einstein made a similar confession: “Imagination is more important than knowledge. Knowledge is limited. Imagination encircles the world.” He wasn’t dismissing science. He was reminding us that disciplined knowledge often lags behind the leap of imagination. First comes the daring sketch in the mind’s eye; only later does proof arrive in the lab.

Jobs’ Calligraphy Class

Steve Jobs, in his well-worn Stanford address, told the story of a calligraphy class he audited in college. No ROI. No plan. Just fascination. A decade later, when building the Macintosh, those lessons surfaced—proportional fonts, beautiful typography on personal computers. A side road became a highway.

Nobody could have drawn a straight line from medieval glassmakers, to Twain’s ticket stubs, to Einstein’s notebooks, to a twenty-year-old in an Oregon classroom doodling letterforms. Yet the pattern is the same: what seems incidental today often becomes indispensable tomorrow.

The Transformer Surprise

Fast forward to 2017. A team of researchers at Google Brain, trying to improve translation models, unveiled something called the transformer architecture. Their ambition wasn’t to remake civilization—it was to help machines handle longer sentences with better accuracy.

What happened next is still unfolding. Transformers turned out to be general-purpose engines of pattern recognition and generation. Within five years, they’d leap from translation into image creation, protein folding, drug discovery, and the large language models behind generative AI. The world shifted—not because anyone set out to build “artificial imagination,” but because a fix for one problem turned out to solve a thousand others.

The Double-Edged Multiplier

Which brings us to now. AI can multiply human creativity—but it can also mute it if we’re not careful. If we only reward what algorithms optimize, we risk sanding off the rough edges where real breakthroughs often hide.

The medieval mason cutting glass wasn’t thinking about infrared filtering. The street artist painting under a highway overpass isn’t running cost-benefit models. The dancer, the sculptor, the kid chalking color onto a sidewalk—none of them are optimizing for quarterly returns. Yet in their raw expressions, new ideas are seeded. Some remain “just art.” Others may one day ripple outward into medicine, architecture, or even the algorithms themselves.

That’s why economies matter here. If all the incentives push toward purpose-first productivity—what AI can scale, what businesses can monetize—we’ll end up with a thinner, quieter world. To keep imagination abundant, we need systems that reward both: the intentional, purpose-driven innovation and the raw, serendipitous creativity that doesn’t yet know its utility.

Why This Matters

Medieval stained glass, Twain’s travels, Einstein’s imagination, Jobs’ calligraphy, transformers for translation: each was pursued for its own sake. Each delivered more than intended.

The lesson? We should value the arts, imagination, and pure R&D not only for what they aim at, but for what they might accidentally unlock. And in this new era, we must be deliberate about pairing AI’s amplifying power with space for human messiness—dance, paint, improv, exploration, drift.

Because sometimes the most important innovations arrive sideways. They aren’t optimized. They’re stumbled upon.

The light that enters a cathedral isn’t just photons through glass. It’s centuries of human curiosity reminding us that secondary effects are often the real breakthroughs—and that the only way to keep discovering them is to keep creating, freely, without knowing in advance where it all leads.

From Matt Gullett at Between Silicon and Soul