Stanford's Erik Brynjolfsson: The Next 10 Years Could Be the Best — or the Worst — in History

Erik Brynjolfsson has spent 30 years measuring what technology does to jobs, and he saw AI coming before almost anyone. So when the Stanford economist says we've "just turned the corner," it's worth listening. In a conversation with Silicon Valley Girl, he lays out where the jobs are already disappearing, why cheaper AI can paradoxically create work, what humans will do for money once intelligence is automated — and why the coming decade could be either the best or the worst in human history.

This post condenses the episode. The research, forecasts, and claims are Brynjolfsson's.

The Canaries Are Already Falling

Brynjolfsson's recent "canaries in the coal mine" paper found something stark: employment is down about 16% for workers under 25 in the most AI-exposed occupations, like coding and call centers. He's careful not to sugarcoat it — the declines are double-digit for people using AI to automate their work — but he insists that's only half the story. At the other end of the spectrum, the least-exposed jobs (home health aides, for instance), older workers, and people using AI to augment rather than replace their work are all seeing growth. And the automation effect, he says, keeps getting bigger every month.

The key, he argues, is to stop thinking about whole jobs and start thinking about tasks. Every occupation is a bundle of them, and there isn't a single one where LLMs "run the table." His favorite example is the radiologist: machines now read medical images, but radiologists perform 26 distinct tasks — physical exams, lab review, coordinating care — most of which remain firmly human. "Everybody's a coder now," he adds, noting that with tools like Replit, Cursor, and Claude Code, every one of his Stanford students shipped running software this year, up from PowerPoint slides the year before.

Why Cheaper AI Can Create Jobs

Here Brynjolfsson gets, in his words, "a little bit wonky" — and it's the most counterintuitive part of the conversation. When AI makes a task cheaper, the effect on employment depends on the demand curve. If demand is inelastic, cheaper output means less total spending and fewer jobs. But when demand is elastic — like air travel after jet engines made flying cheap — lower prices unleash so much more volume that total spending rises. Roughly half the economy works this way. So AI is simultaneously destroying jobs in some corners and creating opportunity in others, and it's the creation half where he spends his energy, through his Stanford course and his startup Work Helix.

He's also candid that the payoff hasn't shown up yet. Benchmark capabilities are skyrocketing while real productivity change stays muted — partly because companies produce "AI slop" or waste the tech on trivia (he recalls a hackathon whose winner used an LLM to generate lunch menus). That gap between capability and impact, he says, is the opportunity. In his PhD work he studied how electric motors took about 30 years to show up in American factory productivity; AI, he predicts, will close its gap in three to five. He's even made a wager with skeptic economist Bob Gordon that productivity by 2030 will blow past official forecasts — because, contrary to the hype-fatigue narrative, he thinks AI is underhyped.

The Future of Work Is Managing Agents

So what does a job look like on the other side? Brynjolfsson frames every project as three parts: defining the question, executing it, and evaluating the result. AI is rapidly mastering the middle, which means most people will spend their time on the first and third — asking the right questions and judging the output. In his telling, everyone becomes something like "the CEO of a fleet of agents." It's a learnable skill, and it rewards a specific blend: pure technical ability misses the real problem, while domain knowledge alone misses where the technology can help. Combine them, and you add the most value.

That has consequences for education. Cookbook courses that teach step-by-step procedures will fade as AI performs them, while the liberal arts — philosophy, art, music appreciation — become more valuable precisely because they cultivate taste and judgment. He half-jokingly suggests universities go back to their liberal-arts roots.

When Silicon Valley Girl reads him a list of well-paid jobs, he doesn't flinch. Junior software engineer, mid-level marketing manager, paralegal — all "in the bullseye," especially anything junior. But radiologist, at $350K, is still a good job: Geoffrey Hinton famously said in 2017 to stop hiring them, and got it badly wrong, because image-reading is only one task and cheaper scans simply generate far more demand.

The Trap Beneath the Pyramid

The disappearance of entry-level roles worries Brynjolfsson at a systemic level. Strip out the base of the corporate pyramid and it becomes a "diamond" — so where do tomorrow's middle managers and senior experts come from? He calls it a coordination problem: it may be individually rational for each company to skip junior hiring, but collectively society needs those people to learn and rise. He praises Infosys for still hiring juniors and explicitly teaching them the big-picture skills they used to absorb "by osmosis." And he draws a hard lesson from globalization: economists (himself included) promised free trade would grow the pie, delivered the disruption, but never delivered the help for those it hurt — producing today's backlash and century-high tariffs. AI, he warns, is "10 times bigger," and the transition needs a plan.

What Humans Are For

All of this sits inside a bigger frame. In his book The Second Machine Age, Brynjolfsson describes the industrial revolution as the moment machines augmented muscle power and made us 30 to 50 times richer — the "first singularity." Now we're augmenting minds, which he believes will be at least as large, faster, and reach deeper into a mostly cognitive economy. Which raises the trillion-dollar question: if intelligence itself is automated, what do humans make money with?

His answer has four parts. Initiative and agency — he insists AI should stand for "Amplifying Intention," useless without a plan but a massive multiplier with one. Human connection — people still play and watch other humans; we'll increasingly prize work that is "certified human" and authentic. Physical trades like plumbing, though he concedes "the window's closing." And, crucially, all the jobs no one has imagined yet — 200 years ago, he notes, no one would have predicted "podcaster." He borrows a line from Reid Hoffman: humanity's superpower is improvisation. Put an ordinary person against the world's best chess computer with their life on the line, and they might just win — not by playing better, but by figuring out some way to short-circuit the machine.

Reinventing Money — and Measuring Value

Brynjolfsson thinks the tools we use to understand the economy are themselves obsolete. Traditional GDP treats zero-price goods — Wikipedia, YouTube, free ChatGPT — as worth nothing, even though people value them enormously. His alternative, GDP-B (B for benefits), asks how much you'd need to be paid to give a product up. Across 600 goods, it reveals trillions in hidden value, with chatbot value alone rising about 70% in nine months. Looking further out, he imagines a world where robots make basic needs essentially free, leaving status — inherently zero-sum — as a key scarce good. The real task of the future economy, he says, is to channel that status competition into something productive: be Einstein, cure diseases, rather than chase status that helps no one. The 21st century, in his view, will need new economic rules the way Adam Smith and Keynes defined earlier ones.

Excited and Scared

Brynjolfsson refuses a tidy ending. The next decade, if we play our cards right, could be "the best decade in human history by far" — unprecedented wealth creation, and longevity gains he ties to Demis Hassabis's hope of curing most diseases within ten years. But it could equally be one of the worst: engineered viruses, AI-manipulated social media, mass concentration of power, autonomous drones already hunting human beings. He's openly worried that the coming wealth will concentrate dangerously, and floats redistribution — universal basic income, wealth taxes — as a backstop his Silicon Valley friends will hate, arguing that even billionaires don't want a world of gated communities and pitchforks.

His deepest point, though, is about agency. More powerful tools, by definition, give humans more power to shape the world — so the right question isn't what AI will do to us, but what we want to use it for. "Be philosophers," he urges. "Think about our values." The doomers aren't wrong that catastrophic risk is real; they're wrong only if they think it's inevitable. As he puts it: "If you're not both excited and scared, you're missing at least half the story."

---

Originally published on Silicon Valley Girl. Watch the full episode: https://www.youtube.com/watch?v=72duHF7iZiU