Michael Darius Eastwood · Independent AI alignment researcher
Published
Willow appears in the book more times than any other single external result. This note collects the passages, states what the book argues from them, and marks where the argument is speculative and the book says so.
Willow is a 105-qubit superconducting processor. Its result is that adding more qubits to the error-correcting code reduces the error rate rather than increasing it, meaning that recursive correction can beat the physical error floor as codes scale. In the field's own terms, this crosses a threshold that had been sought for two decades. The result is a quantum computing milestone; it is not an alignment result and the book is careful about that.
The book makes three arguments from Willow, each stated clearly and each pitched at a different confidence level.
The first argument is thematic: the manuscript that landed the day before Willow argued precisely that correction had to scale with the mechanism it was correcting. Willow demonstrated one instance of that pattern at the quantum level. The book presents this as convergence in shape, not proof of the framework.
The second argument is stronger: Willow provides empirical evidence that recursive self-correction can be built into physical law rather than added on top of it. The book uses this to argue that safety architecture for AI should be architectural rather than peripheral, drawing an analogy between error correction in physics and ethical architecture in AI. This is a strong analogy but not a proof.
The third argument is the most speculative, and the book flags it as such: that the ARC principle's R squared term is validated at one scale by Willow's demonstration that "recursion compounds, that it scales with a squared term because each iteration builds on the previous one." The book says explicitly that this is a scale-jumping analogy, not a proof.
The book acknowledges the Willow warning it "cannot ignore." Quantum computing at scale changes the AI landscape faster than any of the safety machinery the book proposes can be built. "The Quantum Ethical Gates I have proposed are not ready." A quantum path to advanced AI would arrive before the substrate-level architecture the book argues is necessary. The book treats this as a genuine risk, not a rhetorical concession, and returns to it in the timeline chapter.
Willow is a convergence in the register because the shape of the result matches the shape of the earlier manuscript's claim. The register does not credit the manuscript with predicting Willow; the argument is thematic, not causal, and the register's row for Willow is labelled convergent, not prediction.
Three arguments at three confidence levels, one candid warning, and a register entry that credits Google Quantum AI with priority on the result while noting the shape-of-argument convergence with the earlier manuscript. The book uses Willow more than any other external result; it does not overclaim what Willow shows.
From the book Infinite Architects: Intelligence, Recursion, and the Creation of Everything by Michael Darius Eastwood.