The Californian Ideology: Why AI Companies Think the Way They Do, and What Education Can Do About It

 If you want to understand why today's leading AI companies behave the way they do, prioritizing scale and monopoly over caution, treating "intelligence" as a single measurable, ownable quantity, you have to go back further than the founding of OpenAI or DeepMind. You have to go back to a horse farm in 1876.

A fusion of two contradictory instincts

In 1995, media theorists Richard Barbrook and Andy Cameron coined the term "Californian Ideology" to describe a strange fusion running through Silicon Valley: the anti-establishment, information-wants-to-be-free idealism of 1960s counterculture, fused with an uncompromising, winner-take-all capitalism. You can see both strands throughout the region's history, sometimes in the same company, sometimes in open conflict.

The open, communal strand shows up in the "hacker ethic" born at MIT and carried to the Bay Area's Homebrew Computer Club: the belief that knowledge should be shared freely, that you learn by taking things apart, that centralized authority (corporate or governmental) should be distrusted. It's the spirit behind the open architecture of the early Apple II, and it's the direct ancestor of today's open-source software movement.

The competitive, monopolistic strand traces back further, to Leland Stanford's horse-breeding "Palo Alto system": identify superior potential early, and concentrate resources on it rather than letting the market sort things out slowly. That instinct became Stanford University's founding logic, then the venture capital model, then, in psychologist Lewis Terman's hands, the IQ test used to sort children and WWI recruits into tiers of worth. It resurfaces in William Shockley's brain-teaser hiring at Shockley Semiconductor, in PayPal's puzzle-obsessed recruiting decades later, and, most explicitly, in investor Peter Thiel's 2014 book Zero to One, whose central argument is that "competition is for losers" and that the goal of any real company is monopoly.

Why this matters for AI specifically

The reason this history isn't just trivia is that the same instinct, treat intelligence as a single measurable quantity, and bet everything on whoever seems to have the most of it, is baked into how the leading AI labs talk about their own mission. "Artificial general intelligence" is, definitionally, the idea that intelligence is a unified, scalable, ultimately ownable resource, an idea whose intellectual ancestry runs directly through the same Stanford psychology department that produced the Stanford-Binet IQ test in 1916, a test used to justify the coercive sterilization of tens of thousands of Americans deemed insufficiently intelligent, and through the same eugenic worldview that Nazi Germany later cited as partial inspiration for its own racial policy.

To be precise about the claim here: this is not an accusation that AI companies are eugenicist in intent. It's a narrower, historically grounded point: the underlying assumption that "intelligence" is one thing, that it can be measured on a single scale, and that whoever controls the most of it deserves outsized power, is an inherited ideology, not a neutral scientific fact. Sam Altman, OpenAI's CEO, was directly mentored by Peter Thiel. The lineage from Terman's test to Thiel's "build a monopoly" philosophy to today's AI labs racing to build the most capable model isn't a metaphor, it's a documented chain of institutions, funding, and personal mentorship centered on Stanford and its surrounding venture capital ecosystem.

The countercultural, open strand of the Californian Ideology hasn't disappeared, it shows up in open-source AI models and in genuine researcher idealism about broadly beneficial AI. But it coexists uneasily with a business model built on scale, defensibility, and monopoly, and history suggests which strand tends to win once serious money is on the table. Google's founders swore off advertising before caving to it under financial pressure; the same pattern of idealism yielding to monopoly logic is worth watching for in AI.

Counterbalancing it through education

If the problem is a narrow, historically inherited definition of intelligence, treating processing speed and pattern-recognition as the whole of human capability, then part of the counterbalance has to be an education system that does not reproduce that narrowness in how it develops and evaluates students.

Teach the history, not just the technology. Students learning to code or use AI tools should also learn where the underlying assumptions came from: that "smartness" as a single sortable trait is a specific historical construct with a specific, often ugly, history (WWI recruit sorting, school tracking, coercive sterilization), not a neutral fact about how minds work.

Deliberately cultivate what IQ-style metrics don't capture. Creativity, empathy, ethical reasoning, and practical, real-world problem-solving are not soft add-ons to a "real" curriculum built around test scores, they are the capacities most likely to be undervalued by a culture that inherited a narrow definition of intelligence from a testing regime built a century ago. Project-based and interdisciplinary work, collaborative problem-solving, and explicit ethics instruction are ways of building this into a curriculum rather than treating it as extracurricular.

Resist reducing AI literacy to "prompting skill." If a school's whole engagement with AI is teaching students to get better outputs from a model, it risks training the next generation into the same instrumental, competition-is-for-losers mindset critiqued above. AI literacy should include understanding what these systems actually are (statistical pattern-matchers trained on human-produced text and data), what they are not (a unified, general "intelligence"), and who owns and profits from them, so students engage as informed citizens rather than as competitors racing to out-optimize each other with the same tools.

Model the open strand of the Californian Ideology, not just the monopolistic one. Open-source projects, collaborative assessment, and genuinely shared knowledge-building in the classroom are the pedagogical equivalent of the hacker ethic, and they're a deliberate counterweight to a "sort the winners early and bet everything on them" model of education that mirrors Silicon Valley's own logic back at its students.

Where I'd flag uncertainty

The historical through-line, Terman to Shockley to Thiel to Altman, is well documented as a chain of institutions and personal relationships, but the claim that this ideology causally drives specific present-day AI company decisions is more interpretive than the historical facts themselves; reasonable people, including people inside these companies, would contest how much of their strategy is ideology versus straightforward commercial incentive (investors expect returns regardless of anyone's philosophical priors). Treat the history as well-supported and the causal story about "why AI companies behave this way today" as a plausible, argued interpretation rather than a proven mechanism.