AI and Peter Thiel's Stagnation Thesis

Craft on the Other Side of Vibe Coding

We were promised moon colonies and fusion reactors; instead we’ve been handed ever-longer footnotes and thicker journals. Peter Thiel’s barb that “we wanted flying cars and got 140 characters” is a hard-hitting cultural diagnosis. I am myself a product of a culture that mistakes the theater of credentialism for genuine breakthrough.

Now AI arrives like an unannounced auditor, scanning our ledgers of expertise and asking one simple question: which entries actually balance?

Mathematics makes that question painfully clear. Andrew Wiles rose before dawn to complete his proof of Fermat’s Last Theorem, and with a single insight he reshaped number theory, opening entire new branches of research. Meanwhile, dozens of capable mathematicians publish correct but incremental papers on elliptic curves—solid work that earns tenure yet leaves the frontier unmoved. In mathematics a proof either stands or collapses under scrutiny; the gulf between a generational genius and competent scholars is a yawning canyon. That canyon is the power law at work: the difference between the 99th percentile and the 99.9th dwarfs the gap between the 50th and 90th. Excellence isn’t linear but exponential. That’s one reason why, as Peter Thiel says, “competition is for losers.”

Contrast Math with “Area Studies,” where experts spend a decade cataloging oral histories, mapping rituals, and debating ever-finer taxonomies of identity. What would it mean to be ten times as expert in area studies? In fields driven by narrative nuance rather than unambiguous outcomes, AI’s fluency forces us to ask whether a genuine frontier ever existed. Unsurprisingly, the softer the discipline the more politics you find in the discipline. When interpretation reigns, everyone has an incentive to form a clique. The social sciences are perhaps the worst place to be because they are perceived to be scientific and yet most social science studies don’t replicate.

Power laws emerge because early leads compound into ever-growing advantages. A mathematician whose theorem breaks new ground gains funding, collaborators, citation cascades, resources that fuel further breakthroughs. Nobel laureates are cited more not just because their work is superior, but because it’s already visible. In software, the handful who architect kernels and compilers beneath the world’s servers accrue network effects: every fork amplifies their code’s influence, attracting more eyes and contributors. Wherever clear metrics exist, proof validity in math, error rates in manufacturing, uptime percentages in tech, the separation between masters and journeymen only sharpens as the frontier advances.

AI’s incursion doesn’t collapse these chasms; it illuminates them. With a few prompts you can draft passable essays on postcolonial critique, sketch marketing strategies, even generate rudimentary legal briefs. This is a modern miracle, but its also an indictment of the sophistry behind so much white collar work.

The bottom eighty percent of performers in writing and analysis suddenly look like mid-pack humanities PhDs circa 2005. Yet the ceiling remains obstinately high. The world’s most visionary engineers are still sending rockets into orbit and writing compiler optimizations that bend human intuition.

We live in the age of vibe coding, but we mustn’t conflate vibe coding with mastery. The ability to 80/20 your way through unfamiliar terrain is a remarkable feat of pattern recognition. Yet mastery still nests at the far end of the power law. It remains possible to be a ten-times, a hundred-times, even a thousand-times performer: the bridge-builder whose design defies seismic forces, the immunologist whose novel mRNA platform inaugurates a new era of treatments, the classical musician whose improvisation in performance arrests collective breath. Those feats demand depth, tacit knowledge, and stakes where failure is unthinkable.

This is Thiel’s stagnation thesis in microcosm. For decades we mistook credential proliferation for progress. We routed vast talent into narrative-heavy, feedback-light sectors, while the material world idled. Instead of pushing our production-possibilities frontier outward, we became experts at rearranging words within it. AI now serves as an intellectual stress test, revealing which domains rest on firm foundations and which float on simulacra.

Not every field bows equally to the great leveling. Disciplines governed by hard constraints still deliver unambiguous feedback.

In fields built on self-referential complexity, practitioners will watch the tide recede and discover how many PhDs were swimming naked. That reckoning is overdue.