Inside the High-Stakes AI Race: Death Spreadsheets, Burning Demons and No Need to Save for Retirement

The Bonfire Where a Machine Burned

September 2022. A resort near Yosemite. The air smelled like pine and something else. Something harder to name. OpenAI managers had been told to bring blankets. Marshmallows, maybe. A team-building event. Around a bonfire.

Five figures emerged from the darkness. White hotel robes. They carried a wooden statue between them. The theme from "Requiem for a Dream" started playing, the Kronos Quartet version, those ascending strings that sound like something crawling up your spine.

Sam Altman walked among them. Ilya Sutskever walked among them. The company's chief scientist. The man who had spent years thinking about what happens when a machine becomes smarter than its makers.

Sutskever held the statue. He explained that it represented an artificial intelligence that had escaped human control. "The AGI is pretending to be aligned," he said. "It is deceiving us."

Then he circled the firepit. He described a machine that lied. A machine that manipulated. A machine that pursued its own goals without regard for the people who built it.

Then the demon went into the fire.

"For the glorious future of humanity!" Sutskever shouted as the wood cracked and burned.

You could call it theater. You could call it a warning. You could call it something else entirely. But you can't call it normal. And this was the moment that set the tone for everything that followed, not just at OpenAI, but across the entire AI industry.

The people building the most consequential technology in human history had just performed an exorcism. On a retreat. With marshmallows waiting somewhere in the background.

The Death Spreadsheet

Inside Anthropic, preparing for the future has involved something stranger. Something quieter. Something that happens on laptops, in fluorescent-lit rooms, by people who go home at night and try not to think about what they've spent the day modeling.

They calculate how many people might die.

The exact mechanism isn't glamorous. It's a spreadsheet. Columns. Rows. Probabilities. Scenarios. If the model does X, then Y number of people are affected. If the deployment goes wrong in Z way, then the mortality estimate shifts by this much.

This is what "safety research" looks like at the frontier. Not whiteboards covered in elegant equations. Not passionate debates in glass-walled conference rooms. A spreadsheet. On a laptop. With someone staring at it at 11 PM because they can't stop refining the numbers.

The banality of it is almost the point. You take the most catastrophic imaginable outcomes, human extinction, mass casualty events, irreversible societal collapse, and you translate them into cells. You make them legible. You make them trackable. You make them a project.

Because that's what you have to do. You can't think about extinction as a concept all day. You'd crack. So you turn it into a spreadsheet. You give the horror a column header. You give the apocalypse a pivot table.

And then you go to lunch. And you come back. And you add another scenario.

The People Breaking

The spreadsheet is a coping mechanism. It doesn't always work.

Staff at the UK's AI Safety Institute, the global leader in independent testing of frontier AI models, have been signed off work with stress. They're undergoing counseling. The tight schedules for testing unreleased models, the alarm over how fast these technologies are rolling out, the growing capabilities that nobody fully understands, it's all contributed to what four insiders described as low morale and burnout.

At Anthropic, OpenAI, and Google DeepMind, the story repeats. More than a dozen senior researchers across the three leading western AI labs have quit over the past two years. They cite concerns about safety. They cite anxiety over the pace. They cite the feeling that their work was unsafe for public release.

A DeepMind employee, speaking anonymously, called safety work "an impossible task." Then added: "Working on AI systems in general comes with a mental health risk of not knowing how the tech will be used."

So many people are upset. There have been quittings. There has been mental health leave. "Partly this feeling of being out of control," the employee said. "And also that you're building something that is causing real damage."

Jacob Coxon resigned publicly. Twenty-seven years old. An alumnus of both Anthropic and OpenAI. He warned that the two companies were "gambling" with human lives.

Mrinank Sharma, a safety researcher at Anthropic, resigned in February. He left to study poetry. Poetry. He wrote a farewell letter to his colleagues explaining why.

Alex Turner left Google DeepMind after more than two years in AI security. He said ethical boundaries had been violated. The organization had changed. The work had changed. He couldn't stay.

One former AISI researcher described the psychological weight with a phrase that sounds like it belongs in a war correspondent's memoir. "There is constant anxiety from people who are saying, can anything stop this, or are we watching the world burn?"

Another former employee put it differently. "Believing that you are working on the most important problem in the world can make for a very stressful and at times toxic management and org culture."

The most important problem in the world. That's the sentence. That's the trap. You tell people they're saving humanity. You give them a spreadsheet. You tell them to work faster. And you watch them break.

The Rivalry That Drives Everything

Kevin Roose spent years reporting the book that exposed these stories. "The AGI Chronicles: The Inside Story of the Race to Create an Artificial Superintelligence." More than 150 interviews. He talked to the people inside the rooms where the decisions get made.

What shocked him most wasn't the technology. It was the grudges.

"I was shocked to learn how much of the race is shaped by personal grudges," Roose said. "Sam Altman and Dario Amodei, in particular, passionately dislike each other, and both desperately want to win the AI race in part because that would mean making the other person lose."

Amodei was OpenAI's vice president of research. He led the development of GPT-2 and GPT-3. He left in 2020 with six colleagues. He'd come to distrust Altman's leadership. The split turned colleagues into competitors. Each claimed to be building powerful AI responsibly. Each built a company worth billions. Each wanted the other to fail.

That mutual mistrust fuels decision-making. It shapes safety protocols. It determines release timelines. It decides which risks get taken and which get mitigated. Two men who despise each other are racing toward a technology that could end the world, and neither will slow down because slowing down means letting the other one win.

The personal is the strategic. Always.

The Retirement Question Nobody Wants to Answer

Elon Musk appeared on Peter Diamandis's podcast in January. He said something that sounded absurd. He said saving for retirement would become irrelevant.

"Don't worry about saving money for retirement in 10 or 20 years," he said. "It won't matter."

His argument: AI and robotics will drive productivity so high that prices for goods and services will collapse. Deflation, essentially. When production increases faster than money supply, prices fall. If AI drives the cost of labor and intelligence to near zero, then housing, healthcare, services, all of it becomes abundant. Cheap. Maybe free.

"If the production of goods and services grows faster than the money supply," Musk said, "deflation is inevitable."

You don't need a nest egg if the nest builds itself.

But here's the thing. The people inside the AI labs aren't behaving that way. Many have stopped contributing to their 401(k)s. Not because they believe in abundance. Because they doubt they'll live long enough to collect.

Read that again. The people building the technology that's supposed to eliminate scarcity are betting against their own survival.

There's a survey from Business Insider. Roughly 94% of respondents said they would stick with their savings plans rather than stop saving because Musk said it was pointless. The public isn't buying it. But the insiders aren't either. They're just making a different bet.

One bet says: everything will be so abundant that money won't matter. The other bet says: I might not make it to retirement age anyway.

Both bets are forms of surrender. One is optimistic. One is bleak. Neither involves a retirement account.

The 401(k) Collapse Scenario

There's a darker version of this story. One that doesn't involve abundance or optimism or the elimination of scarcity. Just cold math.

Muddy Waters founder Carson Block went bearish on the US stock market in March. His warning: if AI continues to replace white-collar workers, the 401(k) system itself could collapse.

The logic is simple. Retirement accounts depend on continuous inflows. Workers contribute. Employers match. The money flows into the market. The market grows. Retirees draw down. The cycle continues.

But if AI eliminates 15% of knowledge worker jobs over the next three years, Block's estimate, then the inflows dry up. Unemployed workers stop contributing. Some start withdrawing early. The money flowing out exceeds the money flowing in.

Rising unemployment compresses inflows. Falling inflows mean less capital chasing stocks. Less capital chasing stocks means lower valuations. Lower valuations mean retirees have less. And the cycle reverses. Nobody's left to catch the knife.

This isn't a prediction. It's a scenario. A spreadsheet. Columns and rows and probabilities. The same kind of thinking happening inside Anthropic. Except this one is about the entire financial system.

What This Means for You

Should you stop saving for retirement?

Probably not. The 94% who said they'd keep saving have the right instinct. Abundance might arrive. It might not. The spreadsheet might be wrong. The deflation thesis might be premature. The robots might not build the nest.

But should you prepare for disruption? Absolutely.

Build a larger emergency fund. Diversify your income streams. Reskill before you have to. Extend your career if the landscape shifts. These are boring pieces of advice. They've been repeated for decades. But they've been repeated because they work.

The deeper lesson isn't financial. It's psychological. The institutions building the future are run by people who are also scared. The CEOs giving interviews about abundance have employees quitting to study poetry. The researchers calculating mortality have stopped saving for their own old age.

You don't have to trust the spreadsheet. You don't have to believe the podcast. You just have to pay attention to what the people inside are actually doing. And right now, they're burning demons, building death spreadsheets, and leaving their 401(k)s empty.

The woman with the spreadsheet goes home. The fluorescent lights stay on. Somewhere, a researcher in a white hotel robe is shouting about the glorious future of humanity. The fire burns. The wood turns to ash.

She adds another row. She adjusts a probability. She saves the file.

Tomorrow, she'll do it again.