The Building Knew Before You Did

James Bridle published *New Dark Age* in 2018, the year before everything started proving him right in ways that must have felt less like vindication than vertigo. The core thesis — that more information has not produced more understanding, that computation has become an opacity engine rather than an enlightenment machine — was a provocation then. Now it reads like a weather report filed the morning before the storm. Bridle argued that we had built systems whose complexity exceeded our capacity to comprehend them, and that this incomprehension was not a bug but the defining condition of the era. Eight years later, we are living in the condition he described, except the fog is thicker than even he seemed to expect.

The prescience is specific and uncomfortable. Bridle warned about computational solutionism — the belief that sufficiently powerful algorithms could resolve political, ecological, and social crises — well before large language models became the universal solvent Silicon Valley now pours over every problem from grief counseling to military logistics. His insistence that "systemic literacy" mattered more than learning to code anticipated the current hollow ring of every government's STEM pipeline rhetoric, now that we have a generation fluent in Python and no closer to understanding why their rent tripled. His treatment of surveillance as environmental — not a discrete act but an ambient condition, like humidity — predated the normalization of facial recognition in grocery stores and the quiet integration of predictive policing into municipal budgets. On climate, he was early in framing it not as a future risk but as a present epistemological crisis: we had the data and could not act on it, because the systems that generated the data were the same systems generating the paralysis. That diagnosis has aged with grim precision. What he did not anticipate, or at least did not foreground, was the speed at which generative AI would collapse the distinction between information and fabrication entirely. The dark age he described was one of too much data and too little meaning. The one we inhabit adds a third layer: data that performs meaning convincingly while containing none.

The blind spots are characteristic of 2018's critical technology writing. Bridle's framework is overwhelmingly Western and Anglophone, treating the internet as a more or less unified phenomenon rather than the increasingly balkanized set of national infrastructures it has become. He could not have predicted the degree to which China's AI ecosystem would develop along genuinely different epistemological lines, nor the extent to which the EU's regulatory apparatus would create a third distinct information climate. His critique of solutionism, while correct, sometimes assumes a unified "we" confronting these systems — a collective subject that looks increasingly fictional in an era of deepening political fracture. There is also a notable absence: the book says little about labor, about the bodies maintaining the server farms and moderating the content and mining the cobalt. The dark age has janitors. They are not well paid.

Certain passages land differently now. Bridle's call for "new metaphors" to replace the Enlightenment's exhausted vocabulary of clarity and transparency reads less like a theoretical aspiration and more like an operational necessity when the dominant metaphor for AI is a black box that its own creators claim not to understand. His insistence that networked systems distribute agency among human and nonhuman actors — once legible as a nod to Latour and posthumanist theory — now sounds like a plain description of a world where algorithmic recommendation engines shape elections and nobody can locate the decision-maker. The book sits in a lineage that runs from Donna Haraway through Wendy Chun to the current wave of AI ethics writing, but its particular contribution was tonal: it refused optimism and pessimism both, opting instead for a kind of alert bewilderment that now feels like the only honest posture available. It gave its successors permission to say *I don't understand this* without retreating into Luddism.

The question the book raises now, which it could not have raised in 2018: If the new dark age was defined by systems too complex to understand, what do we call the era in which those systems have begun to narrate themselves — fluently, plausibly, and with no obligation to the truth?