In what might be the most accidentally honest post in the history of tech CEO communications, Sam Altman laid out two competing realities in a single viral X post on April 26:
Reality one: “Post-AGI, no one is going to work and the economy is going to collapse.”
Reality two: He’s switching to polyphasic sleep because GPT-5.5 in Codex is so productive that he “can’t afford to be sleeping for such long stretches.”
The post hit 1.2 million views. Nine thousand likes. A thousand replies. And somewhere in there, the entire paradox of 2026’s AI moment, compressed into a single man’s sleep schedule.
The “Last Milestone Before AGI”
Altman called GPT-5.5 “the last major milestone before AGI” — and he might be right. Released April 23, GPT-5.5 represents what OpenAI describes as “a new class of intelligence,” particularly in agentic coding tasks through Codex. People aren’t working less with it. They’re working more. Sixteen-hour days overseeing AI agents. Polyphasic sleep schedules. The hustle isn’t slowing down — it’s accelerating.
Which is precisely what makes Altman’s economy-collapse warning so jarring. If the most powerful person in AI is grinding harder than ever because his own product is that good, what happens when that product makes everyone else’s grind irrelevant?
The Five Principles: A Blueprint or a Press Release?
The same weekend, OpenAI published its five AGI development principles:
- Democratization — Resist power consolidation; ensure key AI decisions are made through democratic processes
- Empowerment — AI should help people achieve their goals, learn more, pursue their dreams
- Universal Prosperity — Global quality-of-life gains, not just profits for frontier labs
- Resilience — Collaborate on risks like cybersecurity and systemic failure
- Adaptability — Iterate based on what we learn
These are, on their face, unobjectionable. Who’s against universal prosperity? Who’d argue for power consolidation? But principles are cheap — infrastructure is expensive, and that’s where the money’s actually going. Google just committed up to $40 billion to Anthropic. Microsoft offered its first-ever voluntary buyouts to redirect headcount toward AI data centers. Snap fired 1,000 people and explicitly blamed AI.
The gap between “democratization” and “a handful of companies spending hundreds of billions to control frontier compute” is the story Altman’s sleep schedule accidentally exposes.
Why the Dissonance Matters
Here’s the uncomfortable truth: both of Altman’s statements can be true simultaneously. Right now, AI amplifies the people who know how to use it — hence the hustle. Soon, it may amplify them out of relevance entirely — hence the collapse warning.
The transition period is where the damage gets done. That’s the window where:
- Companies cut jobs while claiming AI “empowers” the survivors
- Workers grind harder to stay relevant to systems designed to replace them
- Policy debates about robot taxes and four-day workweeks happen at the speed of government (glacial) while the technology moves at the speed of venture capital (ludicrous)
- NZ specifically faces an amplified version of this — a small, export-driven economy where white-collar roles in finance, law, and government are prime AI targets, but the policy infrastructure to manage the transition is still being discussed
Altman has previously advocated for robot taxes, four-day workweeks, and public wealth funds. Those are the right conversations. But they’re conversations happening after GPT-5.5 shipped, not before. The product roadmap moves faster than the policy roadmap, and that asymmetry is structural.
The Real Question
Altman’s polyphasic sleep quip is funny. It’s also a confession. The most AI-empowered person on Earth isn’t using his superintelligent tools to work less. He’s using them to work more. If that’s the best-case scenario — the CEO of the company, with every advantage, still grinding at 3am — what does that tell the rest of us?
The five principles are a start. But principles don’t restructure labour markets. Compute does. Capital does. And right now, both are flowing toward concentration, not democratisation.
🔍 THE BOTTOM LINE
Altman’s viral post wasn’t a contradiction — it was a preview. The AI productivity boom is real, and it’s making the people at the top work harder, not smarter. The economic collapse he warns about isn’t some distant hypothetical; it’s the logical endpoint of a system that eliminates jobs faster than it creates new economic arrangements. OpenAI’s five principles are nice framing. But until “democratization” means something more than “everyone gets a ChatGPT subscription while three companies own the compute,” the hustle-collapse cycle is just going to keep spinning.
