The Man Who Saw It Coming - And What He Missed

THE SIGNAL · ISSUE THREE

Leopold Aschenbrenner predicted the future of AI with eerie accuracy. His most important error tells us everything about why The Human Pioneer exists.

"The future, in fact, is much bigger, darker, and more problematical than anything our bright new technology can dominate."

- Governor Richard D. Lamm (1935-2021)

I worked closely with Gov. Lamm during his 1996 presidential campaign. He was the kind of public servant America rarely produces. He told hard truths before they were fashionable. He saw the shape of things to come when others were still admiring the present. He called himself a futurist. His friends called him Governor Gloom. History keeps proving him right.

I thought of Gov. Lamm often this past week. I was reading and rereading the transcript of a 2024 podcast interview that has quietly become one of the most important documents of our era. The guest was Leopold Aschenbrenner, a 22-year-old former OpenAI researcher. In June 2024, he published a 165-page treatise called Situational Awareness: The Decade Ahead. In it, he predicted with extraordinary precision what artificial intelligence would do to the world, and when.

Two years later, we can grade the exam. The results should stop you cold.

What He Got Right

Start with the numbers, because they are staggering.

Aschenbrenner predicted that total AI investment would approach $1 trillion by 2027, with compute clusters consuming power on the scale of nuclear reactors. In April 2025, McKinsey projected $5.2 trillion in total datacenter capital expenditures through 2030. The five largest hyperscalers, Amazon, Microsoft, Google, Meta, and Oracle, are projected to spend $602 billion in 2026 alone. His "1 gigawatt per cluster by 2026" call hit. His "10 gigawatts by 2028" call is visibly under construction. Stargate, the $500 billion OpenAI-SoftBank-Oracle initiative, is being built in Abilene, Texas at multi-gigawatt scale. The trillion-dollar cluster is no longer a thought experiment.

He predicted that AI would outpace most college graduates by 2025-2026. Confirmed. Current frontier models now outperform the majority of college graduates on standardized tests, coding challenges, medical licensing exams, and legal bar questions. The trajectory he described, from preschooler to PhD in two generational leaps of compute, has tracked almost exactly.

He predicted the discovery of what he called "test-time compute overhang." This was the idea that models could think longer on hard problems rather than simply getting bigger. He wrote this three months before OpenAI launched o1 in September 2024, which did exactly that. Then o3. Then DeepSeek R1. Then Claude's extended thinking mode. Then Gemini's reasoning modes. He named the paradigm before it had a name.

He predicted China as a serious AI competitor, with independent innovation rather than pure theft. DeepSeek's January 2026 release briefly topped the App Store and sent shock waves through the investment community. It made his framing look conservative, not alarmist. He predicted confirmed espionage. In January 2026, Linwei Ding was convicted for stealing Google TPU secrets. He predicted that the US-China competition would increasingly look like an arms race. Both governments now explicitly frame AI as a strategic contest.

He bet his entire net worth on his thesis. His fund returned 2,065% in 2025. The market agreed with him.

His investment firm, Situational Awareness LP, started in September 2024 with $225 million. By the end of 2025, its disclosed US equity book had grown to $5.52 billion, a 22-fold increase in one year. By March 2026, it held $13.68 billion. Aschenbrenner committed essentially his entire net worth to the fund. History rewarded him.

What's Still Coming

Some of his predictions remain unresolved. They are accelerating toward resolution.

AGI by 2027: The question is still open. Six months ago it looked implausible. After recent capability jumps in software engineering tasks, including Anthropic's own June 2026 announcement that Claude now writes more than 80 percent of its own codebase, it has become credible again. The definition of AGI is contested, but the trajectory is not.

"The Project": Aschenbrenner predicted that around 2026-2027, the US government would realize that a nationalized, Manhattan Project-style AI effort was the only viable path. We have not seen full nationalization. But the Commerce Department's June 12 suspension of Fable 5 and Mythos 5 was the first clear signal that Washington is beginning to treat frontier AI as a national security asset requiring government control. The shape of The Project is becoming visible.

Agents as drop-in remote workers: We are at the frontier edge of this now. The 2027 model, Aschenbrenner argued, would not make a software engineer more productive. It would replace the need for one. You would interact with it like a colleague. That moment has not fully arrived. But the distance to it is measured in months, not years.

His Most Important Error

Here is where it gets interesting. Here is where Dick Lamm comes back in.

The analysts who have reviewed Aschenbrenner's predictions two years on have arrived at a consistent verdict: his technical predictions aged brilliantly. His sociological predictions did not. One analyst summarized it this way: he was "accurate on the science, optimistic on diffusion and politics." This is the standard failure mode for technically brilliant forecasters. They see the physics clearly. They underestimate the humans.

Aschenbrenner got the compute curve right. He got the capability jump right. He got the geopolitical competition right. What he did not fully account for was the speed at which intelligence would diffuse, not just accumulate. He predicted that open-source AI would fade, that proprietary algorithms would create a durable American moat, that intelligence would concentrate. Instead, DeepSeek proved that capable AI is becoming cheaper and more widely distributed faster than his framework assumed. Intelligence is not concentrating. It is dispersing.

This is not a small miss. It changes everything downstream of the technical predictions.

He answered what is being built. He left unanswered who we need to become to inhabit it.

Intelligence is dispersing. It is becoming a commodity available to individuals, small nations, rogue actors, and teenagers in their bedrooms. So the questions that matter most are not technical. They are human. They are questions about education, governance, consciousness, character, and civilization. They are questions about what kind of people need to exist on the other side of this transition, and how we build them.

Aschenbrenner wrote 165 pages about the machinery. Almost nothing about the inhabitants.

That is the gap The Human Pioneer exists to fill.

What Dick Lamm Understood

Dick Lamm spent his public life insisting that the future would not take care of itself. He understood something that most optimists miss: that technology accelerates history, but it does not resolve it. The hardest questions, the ones about justice, meaning, governance, and what it means to be human, get harder, not easier, as the machines get smarter.

When he wrote that, "the future is much bigger, darker, and more problematical than anything our bright new technology can dominate," he was not being pessimistic. He was being precise. Technology does not dominate the future. Humans do, or they fail to. The question is always whether we are equal to what we have built.

We are not, right now, equal to what we are building. That is not a counsel of despair. It is the most urgent possible call to action.

Aschenbrenner is right that AGI is coming. He is right that the compute clusters are real, that the capability jumps are real, that the geopolitical stakes are existential. But the future he is describing will be inhabited by parents raising children with no map, by teachers using curricula designed for a world that no longer exists, by governments writing laws for technologies that were unimaginable when the laws were drafted, by citizens who have never been asked to think seriously about what it means to share a civilization with non-human intelligence.

That is the frontier this publication covers. Not the machines. The people.

Three Questions for the Road Ahead

Every issue of The Human Pioneer will return, in one form or another, to three questions that the Aschenbrenner interview makes urgent:

What kind of education do we need to build? If AI outpaces college graduates by 2026, then the industrial-age education model, built to produce reliable workers for predictable jobs, is not just obsolete. It is actively harmful. What replaces it? What does a Pioneer education look like? That is the work of our Pioneer Classroom section, issue by issue.

What governance architecture can hold this? Aschenbrenner predicted "The Project," a government takeover of frontier AI. Whether or not that happens in the form he describes, the deeper question is real: who decides? What are the rules? Who holds the line when the machines get smarter than the people writing the rules? That is the work of our Governance Desk.

Who do we need to become? This is the question no technical paper can answer. It is the question of character, consciousness, and civilization. What are the inner qualities, the wisdom, the adaptability, the groundedness, the moral clarity, that humans will need to navigate the world Aschenbrenner describes? That is the deepest work of this publication.

I called Dick Lamm "Governor Gloom" once, half-joking. He smiled and said: "I'm not gloomy. I'm serious. There's a difference."

Leopold Aschenbrenner is serious too. He saw what was coming with remarkable clarity, built a thesis around it, and staked his fortune on it. He was right about almost everything technical. What he left open is what the machinery does to the people, how the people respond, what kind of civilization emerges on the other side... that is the territory we are here to explore.

The future is, indeed, bigger and darker and more problematical than anything our bright new technology can dominate.

That is precisely why human greatness matters now more than ever.

ABOUT THE AUTHOR

Sean S. McGrath is currently serving as CEO/Team Leader of Keller Williams Realty Spokane. He is a 30-plus year political strategist, former World Affairs Council Fellow, and the author of How to Win (Etcetera Press, 2010). He holds an MA in Diplomacy from Norwich University. He worked closely with Governor Richard D. Lamm during the 1996 presidential campaign. He is the founder of The Human Pioneer, the publication at the intersection of human consciousness, education, governance, and the AI and space frontier. Subscribe and join the conversation at thehumanpioneer.com.

thehumanpioneer.com · Human Greatness in the Age of AI