Sequoia Partner Konstantine Buhler: The Cognitive Revolution Has Begun
Sequoia partner Konstantine Buhler stated that humanity has outsourced muscle to build the industrial world and is now outsourcing cognition; he commutes daily using Waymo, where the 5,000-pound vehicle performs physical tasks while predicting road conditions involves cognitive tasks. He cited that the severe injury collision rate of autonomous vehicles is significantly lower than that of human drivers, viewing this as a microcosm of the physical revolution after two hundred years and the cognitive revolution just beginning.
He breaks down work into physical and cognitive: the former is force multiplied by distance, the latter is thinking; both are scarce, have a price, and when prices collapse, supply floods in and demand expands. The physical side has transitioned from steam, internal combustion to electric motors over about two hundred years, shifting from almost entirely relying on biological muscle power to almost entirely relying on machines; the cognitive side has evolved from clocks and Pascal calculators to electronic computing and neural networks. He estimates that the global economy of about $120 trillion can be roughly divided in half by work nature, with the physical half having been mechanized for two hundred years, while the cognitive half has barely begun and will be larger and faster. The industrial era expanded coal, iron, oil, and steel along with converter processes, while today it corresponds to electricity, computing power, and Transformers, with Nvidia, TSMC, model labs, and hyperscale clouds paralleling Carnegie, Rockefeller, and Bessemer. He noted that the cost of intelligence per watt has decreased more than tenfold each year, and cognition will follow Jevons' Paradox: efficiency improvements do not reduce consumption but rather amplify it.
He likened AI programming to the Jenny spinning machine: skilled workers have shifted from hand-spinning to overseeing machines; automotive-grade applications have yet to emerge, with candidates being scientific machine speed, personal assistant-style agents, or new forms of human collaboration. U.S. agricultural employment dropped from over 75% in 1800 to about 1%, while total employment and real wages have risen over the long term; he referenced economic historian Robert Allen's Engels Pause: from 1780 to 1840, British labor productivity rose by 46%, while real wages only increased by 12%, and from 1840 to 1900, productivity rose by 90%, and wages increased by 123%. The Luddites viewed their own lives positively but were wrong about their descendants. He quoted Mark Zuckerberg: there are no rules stating that automation must outpace the expansion of personal capabilities and new skill demands. The speed is the difference: farm children had about forty years to transition to factory labor, while this round may only take five years. Benjamin Bloom's 1984 one-on-one mastery learning allowed average students to outperform 98% of conventional classrooms; he claims that blended human-machine tutoring is more effective and cheaper. In medicine, he pointed out that AlphaFold predicts almost all known protein structures, making long-tail diseases accessible after the collapse of investigation costs. Deep Blue's victory over Garry Kasparov in 1997 and AlphaGo's novelty in professional Go have been used by him to illustrate that human competition has not disappeared. Closing statement: machines can draft treaties, but humans still need to sign them.
Who is buying and who is selling: the event is driven by framing AI as the narrative of the second industrial revolution, not by new employment statistics landing on the same day. The buyers are growth funds defending capital expenditures for computing power, while the sellers are infrastructure companies treating the power grid and wafers as the new steel. Beneficiaries are model labs and foundries, while the education and retraining systems designed by the old clock are under pressure. Funds flow from the declining unit price of cognition to larger-scale clusters, with demand expanding in a Jevons-like manner, rather than replacing jobs one by one.
Source: Public information
ABAB AI Insight
Sequoia frames commuting robots as a part of civilization history to shift anxiety from singularity slogans back to prices and supply. The core equation is: once cognition becomes cheap, previously unthought-of problems will turn into sellable services. Recent safety comparisons of Waymo show severe injuries reduced by over 90%, comparable to the 17-fold difference mentioned, proving that cognitive outsourcing is already happening, not science fiction.
The capital path is written in the triad of electricity, silicon, and algorithms. Money flows from trillion-dollar capital expenditures in model labs and hyperscale clouds into power grid capacity and intelligent output, with state power shifting from refining capacity to grid connection capability. Programming agents are the first chargeable skilled cognition, corresponding to spinning rather than automobiles; funds need to find categories that fundamentally do not exist when thinking is scarce.
Comparative objects range from the Jenny spinning machine to automobiles, from Engels' pause to wages catching up with productivity, from factory-style schools to one-on-one tutors. The industry position is in the spinning phase of cognitive mechanization, with expansion relying on collapsing unit prices, and control lying in computing power and the power grid, rather than the superintelligence timelines in slogans.
Structural change is technological substitution. The mechanism is that the physical revolution sent children from fields to classrooms; if the cognitive revolution takes hold, it will render factory-style schools obsolete, akin to hand-spinning training; whether a pause reoccurs depends on whether retraining timelines can be compressed from forty years to five, rather than whether models claim to be general-purpose.
ABAB News · Cognitive Law
When prices collapse, demand does not conserve but instead is used for things previously unimagined.
The first spinning machine was not an automobile, and the first generation of programming agents is not the endgame.
Machines can draft, but the signing authority remains the trust between people.