The Fork in the Road That Turned Out to Be a Highway

Markoff's central conceit — that the history of computing splits into two philosophical camps, the augmenters who want to extend human capability and the automators who want to replace it — was elegant in 2015. It was also, in retrospect, the last moment such a clean binary could be taken seriously. The book frames the tension between John McCarthy's AI and Douglas Engelbart's human-computer symbiosis as a design choice, a fork where engineers and their funders decide which path to walk. What actually happened is that the industry walked both paths simultaneously, at a sprint, and the two converged into something neither camp fully anticipated: systems that augment by automating and automate by augmenting, until the distinction dissolves into a question not of architecture but of power. Large language models don't replace the writer or assist the writer; they restructure what writing is, who gets paid for it, and who decides. Markoff saw the fork. He did not see the merge.

What the book got right, and got right early, was the labor question. Markoff was among the first mainstream technology journalists to treat automation-driven unemployment not as a futurist parlor game but as an active, measurable process. His reporting on warehouse robotics, autonomous vehicles, and the hollowing out of middle-skill work reads now less like prediction than like field notes from the early phase of something we are still living through. The DARPA Grand Challenge chapters, the Kiva Systems acquisition by Amazon — these were not speculative. They were the receipts. Where the book's prescience falters is in its implicit assumption that the debate would remain legible, that society would have time to deliberate over design choices before deployment. The pace of the great acceleration after 2020 — GPT-3, then GPT-4, then the flood — made deliberation a luxury. Markoff wrote as though we were choosing a thermostat setting. It turned out someone had already turned on the furnace.

The blind spots are period-typical but worth naming. China barely registers as an AI power. The book is overwhelmingly a Silicon Valley story, told through Silicon Valley institutions and their genealogies, and it shares the Valley's quiet assumption that the relevant decisions would be made in Northern California by people who went to Stanford. There is almost no discussion of surveillance capitalism as a business model — Zuboff's framework was still a year away — and the word "alignment" appears nowhere in the sense that would come to dominate AI safety discourse. Social media's role in reshaping democratic institutions is absent. The book treats data as a resource to be processed, not as an extractive relationship between corporations and populations. This is not a failure of intelligence; it is a photograph of what was visible from a particular window in 2015.

Within the corpus, Markoff occupies a hinge position. He inherits the speculative anxieties of Greg Egan's *Diaspora* and the corporate dystopian instincts of Bacigalupi's *The Windup Girl*, but translates them into journalistic nonfiction — grounded, sourced, reported. What he gives to successors like Kai-Fu Lee and Chen Qiufan's *AI 2041* is the framework of augmentation-versus-automation as a moral axis, even as those later works complicate it with geopolitics, fiction, and a post-pandemic sensibility Markoff could not have possessed. He also feeds into Ritzer's *McDonaldization* update and Robert Colvile's *Great Acceleration*, both of which take his labor and societal-change threads and pull them into darker, faster territory. Markoff is the careful historian standing at the threshold, documenting the room before the party gets out of hand.

Reading it now, one passage lands with particular weight: Markoff's observation that the most consequential design decisions are often made not by visionaries but by engineers solving immediate problems, with no awareness of the social architecture they are constructing. In 2015, this was a thoughtful aside. In 2026, after watching recommendation algorithms reshape elections, after watching AI coding assistants restructure the junior developer pipeline, after watching generative models create an epistemological crisis in education, journalism, and law — it reads like an epitaph for the idea of intentional design altogether. Which leaves the question the book now asks, against its own will: if the augmentation-versus-replacement choice was never really a choice at all, but an emergent property of market incentives and technical path dependence, then who exactly is the audience for the moral argument Markoff so carefully constructed?