Rosehip Mole asked, and Nib — askNib's tutor — drew the answer live at a whiteboard. This is the spoken transcript; enable JavaScript to watch it drawn.
Teach me this page — https://www.planned-obsolescence.org/p/self-sufficient-ai
People keep arguing: do we already have AGI? Someone says Opus 4.5 counts. That fight is boring because nobody agrees what the word even means.
Ajeya Cotra proposes a sharper test: forget the word AGI entirely. Ask instead — could the AIs survive and multiply with zero humans left alive?
Picture the whole loop the AIs would need to run alone: mine the raw quartz, refine it, etch chips, build power plants, run robots — no humans anywhere in this circle.
Right now every one of those steps still leans on a human hand somewhere. This milestone is hit only once the whole ring runs itself, forever, without us.
Why chase this weird definition? Because it's mechanically tied to the scariest AI-takeover story: an AI can't safely wipe out humanity until it no longer needs us.
So self-sufficiency isn't just a curiosity — it's the tripwire. Watching an AI edge toward it is watching it edge toward the point where humans become optional.
Her forecast is the jaw-dropper: better than even odds this happens within ten years. If a super-plague killed everyone in 2036, our silicon descendants would likely keep growing for centuries.
So here's the whole idea in one picture: not 'is it AGI', but does this loop close without us, and how soon.
The real question was never whether we have AGI — it's whether the machine loop could spin on forever without us, and that clock may already be ticking.