A programmer, upset that Amazon had turned his Fire tablet into a stupid piece of plastic and silicon by refusing to support it, turned AI loose on the problem. Around $260 in tokens later, the device is fully jailbroken and fully his own.
In another example, from the New Yorker: off-patent drugs can now be systematically examined by AI for what they might treat - diseases nobody ever tested them against, because nobody could profit from the answer.
The human labour required made each task cost-prohibitive; the collapse in the cost of cognitive labour has changed that calculus.
Both tasks are economic orphans. Amazon abandoned the device; the patent system abandoned the drugs. In both cases the value is real, but it belongs to no one who could pay to extract it - so under the old economics of expensive cognition, it simply sat there. Unowned value, unharvested. Forever.
A cognition surplus doesn't merely make existing work cheaper. It makes orphaned value harvestable for the first time - the enormous inventory of tasks that are worth doing but structurally unprofitable: liberating abandoned devices, repurposing abandoned molecules, transcribing family archives, maintaining unglamorous open source, making old documents accessible, translating work no publisher will touch. The market never funded these, by design. The surplus doesn't ask the market, doesn't need the market. It's the first cognitive resource in history with no opportunity cost, and value nobody can own is exactly what it's for.
These are the weak early signals of a phenomenon that will mushroom over the next 18 to 24 months, as good-enough AI arrives everywhere - in our businesses, in our homes, for every class of task we deem worthwhile.
Sixteen years ago Clay Shirky described the cognitive surplus: the world's educated free time, a trillion idle hours a year, which - once the network aggregated it - built Wikipedia in its spare moments. Before that, millions of households donated their computers' idle cycles to
SETI@homeand Folding@home: spare compute, pooled into public goods. In both examples, idle capacity found its shape. But both times the surplus had to leave home to be useful - uploaded, aggregated, centrally directed. This surplus is different. The cycles think now. They don't need a project to report in to. Every household can run its own Folding@home, on problems it chooses, for beneficiaries it names - starting with itself.
We are not used to thinking in these terms, backed by an army of cognition: always on, always ready, never taking time off or a sick day. Everything we've been taught to fear about AI - the tirelessness, the constancy, the patience - turns out, in our own hands and to our own ends, to be the entire benefit.
At the moment, we can't imagine a world of low-to-no-cost cognition - and so we let nearly all of it idle, wasted, the way we once let a trillion hours of human attention drain into the television. Flipping the script, we get to re-imagine what we can do with the cognition surplus, developing both techniques and mindset that accompany this shift.
At the very least, we should be asking the cognition surplus itself how it might best be put to work. Those of us already running agents through the night know what happens next: asked, they answer - with plans, workflows, and questions of their own. The first assignment of the cognition surplus is the job of imagining its second.
Maybe that's where this begins: with an inquiry to the surplus. It may have some thoughts. Perhaps it can inspire a few of our own.
This will take some getting used to.