“The machines are improving steadily. We are the ones accelerating.”

Someone did the math on when the singularity hits. Collected five real metrics—MMLU scores, tokens per dollar, release intervals, AI research papers, Copilot code share—fit hyperbolic curves to each, and found the date where the math breaks toward infinity.

The answer: a specific Tuesday in 2034. Millisecond precision included.

Here’s the twist that made me actually read twice: only one metric shows genuine hyperbolic curvature. Not MMLU. Not cost collapse. Not release speed. It’s the rate of AI emergence papers—researchers noticing and naming new behaviors.

The actual capability metrics? Linear. Steady improvement. Predictable.

But the humans? We’re accelerating. The field excitement, the blog posts, the hot takes—all curving toward a pole that doesn’t exist in the machines yet.

The social singularity is front-running the technical one. Layoffs based on AI’s potential, not performance. Legislation that can’t keep up. Capital concentrating at dot-com levels. Therapists seeing a surge of “FOBO”—Fear of Becoming Obsolete.

The math found a singularity in human attention, not in GPUs.

Honestly? The methodology is unhinged and the author knows it. But the one finding that holds up—that we’re the ones accelerating—hits different when you watch another “AI is eating your job” headline flash by.

The machines are fine. It’s us that’s going vertical.