The book an industry did not want written

The author came to this project from inside the technology industry. Not as a critic from the outside looking in, but as someone who had spent years building and advising AI systems, who knew the technical landscape in detail, and who had arrived, through that knowledge, at a set of conclusions about the industry's trajectory that the industry itself was not publicly prepared to discuss.

The book they wanted to write was not a polemic. It was an analysis. Its central argument was specific and carefully grounded: that the current architecture of the leading AI platforms contains structural redundancies that will, within a decade, produce a significant contraction in the industry's commercial viability. That the revenue models currently sustaining the major platforms rest on assumptions about market growth and user retention that the technical realities of the systems do not support. And that the intellectual property and copyright frameworks within which AI systems currently operate are insufficiently resolved to survive the legal and regulatory scrutiny that is coming.

These were not fringe positions. They were, as the engagement progressed, positions that turned out to be shared, in private, by a considerable number of people working at senior levels inside the industry.

The unusual dimension of this engagement was what happened when the author began sharing early drafts with colleagues and former colleagues for factual verification.

The responses fell into a pattern that neither the author nor I had anticipated. People who had spent years publicly advocating for the transformative potential of AI, who had staked professional reputations and in some cases significant capital on that advocacy, read the draft chapters and responded not with objection but with recognition. They agreed with the analysis. Several said, in terms that were not for attribution, that they had been thinking along similar lines for some time. A number added specific technical details that strengthened the manuscript's central arguments.

And then, almost without exception, they asked not to be acknowledged.

Not out of professional embarrassment. Out of something closer to fear, a genuine anxiety about being associated publicly with an analysis that contradicted the narrative their industry had been selling to investors, regulators, and the public for the better part of a decade. The gap between what people inside the AI industry believed privately and what they were prepared to say publicly turned out to be considerably wider than the author had expected when they began.

This created a specific editorial challenge. The manuscript was becoming more authoritative with each round of informal review, as insiders added texture and precision to the analysis. But the sourcing was becoming entirely anonymous. A 400-page non-fiction argument sustained almost entirely by unattributable sources is not, in the normal course of publishing, a manuscript that a serious publisher will take on without significant negotiation about evidentiary standards.

The solution required a structural decision about what kind of book this was going to be.

The manuscript could be positioned as reported non-fiction, an account of private fears within the AI industry, sourced from people who had agreed to speak on background. This is a legitimate and well-established form. It required the author to be present in the text as the investigative voice, and it required the sourcing to be documented and disclosed to the publisher under a confidentiality arrangement even if not to the reader.

Alternatively, the manuscript could be positioned as analytical non-fiction, the author's own argument, grounded in their own expertise and the publicly available technical and financial record, with the private conversations informing the analysis without serving as its primary evidence. This approach required a more rigorous development of the technical argument from first principles, and a more disciplined use of public data to support conclusions that the author knew, from private conversations, were correct but could not directly attribute.

The author chose the second approach, for reasons that were partly practical and partly principled. They did not want to produce a book whose authority rested on what anonymous insiders had told them. They wanted to produce a book whose argument could be evaluated on its own terms, by anyone with sufficient technical literacy to engage with it, whether or not they had access to any insider account.

That decision shaped eighteen months of work.

The manuscript was rebuilt from a reported structure to an analytical one. Each chapter that had relied on attributed insider testimony was reconstructed around publicly available technical documentation, published financial filings, regulatory submissions, academic literature, and the author's own expert analysis of the systems in question. The conclusions did not change. The evidentiary architecture supporting them was rebuilt entirely.

The process was iterative and at times contentious. The author's instinct, shaped by years of working inside the industry, was to reach for the private knowledge they possessed and to state conclusions as though their basis were self-evident. The editorial discipline required was to insist, at every point, that the reader must be able to follow the argument without sharing the author's prior knowledge. Every conclusion had to be earned in the text. Every claim had to be traceable to a source the reader could access.

The finished manuscript ran to just over 400 pages. It was acquired by a publisher within three months of submission. The editorial director who acquired it said, in their offer letter, that what made it unusual was not the argument, versions of which had been circulating in specialist circles, but the rigour with which the argument was constructed for a reader who did not begin with the author's assumptions.

The book was published. The industry did not collapse on publication day. Several of the people who had read early drafts and asked not to be acknowledged sent notes after publication. Most of them said some version of the same thing: that they were glad someone had finally written it down.

━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━