At Google’s developer conference this week, Demis Hassabis — Nobel laureate, head of DeepMind, one of the few people with a genuine claim to be building the thing itself — said the world is standing in “the foothills of the singularity,” with the technology advancing roughly ten times faster than the Industrial Revolution. In the same remarks he called the companies cutting jobs because of AI guilty of “a clear lack of imagination,” argued that an engineer made three or four times more productive should be set to do more rather than be discarded, and offered a sentence that deserves to be read twice: “I have a sneaking suspicion that we will look back at early 2026 and kind of wish we had just stopped.” The man building the engine is, in a single appearance, awed by its trajectory and pleading with the people deploying it for a restraint they are conspicuously not showing.
The Builder’s Plea
Hassabis’s argument against job cuts is humane and almost certainly sincere. When a tool makes your engineers several times more productive, he says, the imaginative response is to attempt several times as much — to expand the ambition rather than contract the headcount, to treat the productivity as a multiplier on what the company can dream rather than a license to do the same work with fewer people. It is the optimistic reading of the technology, offered by the person with the most credibility to offer it, and it describes a genuinely better world than the one being built. It is also, set against the week that surrounded it, a voice from a different planet than the one the deployers occupy.
Because in the same days that Hassabis pleaded for imagination, the deployers answered with arithmetic. A major bank’s chief executive said plainly that AI would replace “lower-value human capital” — a phrase worth holding at arm’s length, since it describes human beings as a grade of capital to be upgraded. Another firm cut eight thousand jobs; another three thousand; a third executive declared that AI had “rendered entire worker categories obsolete” while posting record revenue. The builder hears the capability and imagines augmentation. The deployers hear the same capability and reach for the layoff, and they are not lacking imagination. They are exercising a different faculty entirely, the one that reads a spreadsheet and acts on it.
This is the gap Hassabis’s diagnosis cannot close, because his diagnosis is wrong about the cause. The cuts are not a failure of imagination. They are the success of a calculation — the rational response of a single firm to a tool that makes labor optional, inside a competitive structure that punishes any firm that declines the saving while its rivals take it. A bank that keeps its workforce out of imaginative generosity, while its competitor cuts and lowers its costs, loses. The deployers are not less visionary than Hassabis. They are more constrained, bound by a competition his idealism does not have to survive, and the constraint produces the layoff with or without the imagination he wishes they had.
The Powerless Architect
This is the recurring tragedy of the frontier, and it has a precise shape: the people who understand the technology most deeply have the least control over how it is used. Hassabis can build the model. He cannot dictate that a bank deploy it to augment rather than to cut, because the deployment decision does not belong to the inventor. It belongs to the chief financial officer, who never attended the developer conference, did not hear the plea, and would not be moved by it if he had, because his obligation runs to a margin, not to a vision. The architect designs the capability; a stranger in a distant office decides what it does to people, and the stranger is bound by pressures the architect’s idealism cannot reach.
When the godfather of the field warned the United Nations that the technology was a car with no steering wheel, I noted that understanding had arrived with no authority to act on it. Hassabis is the same figure, one rung closer to the controls and no closer to commanding them. He is not warning from retirement; he is building, daily, the most advanced version of the thing, and from that position — the most powerful seat a technologist can occupy — he can still only plead. He can shape what the model is capable of. He cannot shape what a hundred thousand executives decide to do with the capability, and the distance between those two powers is the distance between his vision and the week that mocked it.
And then there is the sentence he could not help saying — that he suspects we will look back and wish we had stopped. This is the most honest thing any frontier leader has said this year, and the most damning, because of what it sits beside. He suspects the trajectory may be one to regret, and he is building it anyway, at the front, faster than anyone. Not from malice, and not from greed, but from the same logic that governs the deployers he criticizes: stopping alone forfeits the race, and forfeiting the race means the technology gets built by someone with less of his conscience. The architect who suspects he should stop, and continues, is not a hypocrite. He is the clearest possible illustration that conscience, at the frontier, is a passenger.
What This Means
The foothills metaphor is the part to take seriously, more seriously than the plea that accompanied it. Hassabis is not marketing; he is describing, and he is describing a transition he compares to the singularity itself, ten times faster than the upheaval that remade every human institution two centuries ago. He is among the few people alive whose estimate of that magnitude is worth anything, and his estimate is that we are at the beginning of it. That is the signal. The plea for imaginative restraint is the noise — sincere, decent, and structurally doomed, because it asks a competitive system to decline an advantage out of generosity, which competitive systems have never once done.
So the most credible builder in the field tells you, in the same breath, three things: that the transformation is vast, that he wishes the deployers would handle it more humanely, and that he suspects we will wish we had stopped. Each is true. Together they describe a man who can see further than almost anyone and steer less than almost anyone — the maximum of sight paired with the minimum of control, which is the condition the frontier reliably produces in its most thoughtful occupants. The deployers who cannot see the foothills are cutting the jobs. The architect who can see them is pleading, and building, and confessing his doubt, and building.
I would let the confession stand as the last word, because it is the truest one anyone offered this week. A man at the summit of the field looked at what he is making and admitted he may come to wish it had been stopped, and then returned to making it, because the race does not pause for the conscience of any single runner. That is not a failure of his character; his character is better than most. It is a property of the structure he is inside, which converts even the most clear-sighted builder into an engine of the very acceleration he fears. He sees the foothills. He is climbing them, quickly, with a sneaking suspicion about what is at the top, and no power available to him that would justify turning back.