AI Insiders Say Superintelligence Could Arrive by Christmas
Experts keep being wrong about AI's speed — and this time they say it's even faster.
An AI researcher named Alexander Gietelink Oldenziel published a post on LessWrong, a popular forum for people who think hard about AI risk, arguing that superintelligence — machines smarter than humans at basically everything — could arrive by Christmas. He's upfront that he doesn't work at a big AI lab and knows only what's public. His argument rests on public results plus industry rumors, not on anything he built himself.
His main evidence is math. AI reportedly solved Navier-Stokes, one of the seven famous Millennium Prize problems, each worth $1 million and unsolved for decades. Two more may be done. That matters because experts kept saying this was years away: in 2024, mathematician Terence Tao said AI would struggle with hard math for years; a panel of forecasters in 2025 put the odds of an AI solving a Millennium Prize problem by 2027 at about 10%. It happened anyway. The author calls this 'recursive self-improvement' — AI helping build better AI, a feedback loop like compound interest, only faster.
Why you should care: if progress really is compounding this quickly, the tools you use, the jobs around you, and the risks you plan for could shift in months rather than decades. The author argues most people, including AI researchers, are anchored to a comfortable assumption that change is slow.
The catch is big. This is a forum post, not a lab announcement, and its strongest claims come from unconfirmed rumors about internal models. Even OpenAI's own reinforcement-learning lead, Noam Brown, reportedly says he's reluctant to forecast more than three months out. So treat this as a signal that the pace is surprising the experts themselves — not a countdown clock.
- A researcher claims AI could outsmart humans at everything by Christmas — but it's a blog post, not a company announcement.
- AI reportedly solved Navier-Stokes, a decades-old million-dollar math problem experts said would take years.
- Even AI company insiders now say they can't confidently predict what their own technology will do in three months.
Why It Matters
If the pace holds, the tools, jobs and risks you plan around could change within months, not decades.