Diomidis Spinellis — a computer scientist who has spent decades making a living from thinking — writes that he feels like he's digging his own grave. His thesis is blunt: AI systems are improving across perception, reasoning, judgement, and autonomous action so rapidly that the bulk of today's cognitive work will be automated. Three drivers — better performance, lower cost, higher availability — make the trajectory structurally inevitable regardless of regulatory resistance. The performance claim is anchored to a specific arc. From ChatGPT's 2022 launch, which couldn't count the letters in his name, to OpenAI's release of math papers where an internal AI model solved many hundreds of previously open math problems. That's not incremental improvement on benchmarks — it's category-crossing capability growth in under four years. Spinellis pairs this with a darker observation: human proficiency in key information-processing skills is not merely stagnant but actively declining among students. The cost curve is equally directional. Successive generations of AI models offer superior performance at lower cost, with older models retained only for backward compatibility. Trillions of dollars in projected data centre investment through 2030 will bring vast additional processing capacity online. Each dollar of AI compute gets cheaper; each hour of human cognitive labour does not. The financial logic for substitution widens every quarter. Spinellis predicts fiercer resistance than prior technological revolutions precisely because the displaced class — knowledge workers — holds greater economic and organisational power than factory workers ever did. Their specialised expertise, control over information, and decision-making authority give them political leverage to demand regulation. But he dismisses this as a holding action: governments will try to clip AI's wings through regulation and taxation, and such actions didn't quell past revolutions. The most striking move in the essay is the historical analogy: thought will increasingly become a leisure pursuit, as hunting did following animal domestication, calligraphy after typography, woodworking after industrialisation, and horse riding after motorisation. This is not a prediction about unemployment — it's a prediction about the decoupling of cognition from economic purpose. The claim is that thinking will persist as a human activity but lose its market value. What's conspicuously absent is any structural analysis of who captures the surplus. If AI replaces cognitive labour, the productivity gains accrue somewhere — and the essay doesn't ask where. The advice offered — keep pace with AI advances, be flexible and resilient, build a purposeful life beyond work — is individually sensible and collectively insufficient. It treats the transition as a personal adaptation challenge rather than a distributional one. The essay is honest about the direction of travel but silent on the political economy. Revolutions don't just transform production — they redistribute power. The question isn't whether thinking becomes a hobby. It's whether the billions of people whose livelihoods depend on cognitive work get a seat at the table where the new surplus is divided, or whether the analogy that actually fits is not horse riding but horse glue.
