AI’s Failure of Imagination
We often think the origin of the term “Failure of Imagination” is from that horrendous September day of 2001. In fact the term and the concept is much older with roots back the Historian David Hackett Fischer who coined the phrase prominently in Historians' Fallacies (1970) to describe the logical error of assuming that because you can't imagine an alternative explanation, none exists.
The concept was heavily applied to Pearl Harbor. Roberta Wohlstetter's classic Pearl Harbor: Warning and Decision (1962) essentially argued that the surprise attack stemmed from an inability to imagine what the enemy might do — signals were available, but analysts couldn't conceive of the scenario. This made "imagination failure" a standard term in intelligence analysis literature long before that horrific sun drenched day of 2001.
The term was again applied to the Columbia Accident; the investigation Board (2003) described organizational and cultural failures at NASA in similar terms, and writers like Diane Vaughan applied the same concept to Challenger.
By the 2008 Financial crisis the term had hit the mainstream in post-portems. The single most famous usage: After Queen Elizabeth asked economists at the LSE in November 2008 "Why did nobody see it coming?", the British Academy convened 33 of Britain's top economists and regulators and sent her a formal answer in July 2009. Its conclusion: the failure was "principally a failure of the collective imagination of many bright people, both in this country and internationally, to understand the risks to the system as a whole". In Bernanke, Geithners & Paulson’s Firefighting (2019); all three architects of the response conceded that “these failures of anticipation were in part a failure of imagination”.
The dot-com bubble coined its own vernacular to describe what went wrong after the collapse with terms such as:
"Irrational exuberance" — Greenspan's December 1996 phrase became the era's defining label, framing the bubble as emotional excess rather than a failure to foresee .
"Herd mentality," "groupthink," "mania," "the New Economy paradigm" — the post-crash analysis leaned on crowd-psychology language inherited from Kindleberger's Manias, Panics, and Crashes and Galbraith.
The distinction makes sense: the dot-com crash was widely seen as a predictable bubble deflating — plenty of people imagined it, they just couldn't time it. Nobody called it a failure of imagination because imagination wasn't the problem; restraint was.
But, ironically a British Academy letter itself acknowledged that the dot-com crash helped cause the 2008 imagination failure: because the Fed had successfully "mopped up" after the 2000 bubble burst with only a mild recession, policymakers concluded bubbles could always be cleaned up after the fact . In other words, surviving the dot-com crash without catastrophe convinced the smart money that systemic collapse was literally unimaginable — which is precisely the failure mode the phrase describes.
Ok. So now let’s have a bit of fun. My thinking has always been from First Principles. I analyse situations by identifying the basics, check to see if fundamentals have been identified, is their alignment amongst stakeholders, what structure is present, if any. And, lastly, what and when has any kind of execution been performed in service of the former. It just might be, maybe, if one steps back and runs this seemingly simple analysis the failures of imagination, irrational exuberance, herd mentality, group think and mania can be supplanted with more logical insights to avoid a whole lot of hurt in time, Capital and mental health. What emerges typically are insights of logic and actionable intelligence.
Consider where we sit in August of 2026 with the wonderful and challenging world of AI. Frankly, it’s a total shit-show. And, I’ll examine why. Let's “follow the money” first.
Exuberance (rational or not)-
Six years in and there are roughly $3T, that’s right TRILLION DOLLARS, in commitments. Let me write the number out; $3,000,000,000,000. Put another way, 3T seconds is 95,000 years.
On that $3T, revenues are reported in 2026 at Open AI at $22B annualized revenue, Anthropic reported at ~$25B, XAi at ~$4B, and some smaller players who in aggregate generate ~$1.1. Meta and Google conspicuously do not break out those numbers. In total ~$52B.
Taken together, the frontier labs spend roughly $2 for every $1 generated. Add in Capex costs and that ratio of 2:1 cost to revenue generation, jumps to 5:1. At 5:1, it costs the labs on average $5 to generate $1 of revenue.
The New Economy paradigm-
The demand signal is manufactured. Nvidia is assembling a deal web potentially exceeding $750B; backstopping SoftBank's Ohio campus so OpenAI can lease against Nvidia's credit, financing OpenAI's chip purchases, holding $30B in equity. Google is backstopping Anthropic's datacenter leases . The seller is financing the buyer. Revenue on one side is capex on the other; that's a loop, not a market.
The accounting assumes the future. Hyperscalers depreciate GPUs over five-to-six years when the economic life looks like two or three — Burry's estimate: ~$176B of understated depreciation across 2026–28 . Amazon shortened its depreciation schedule and took the $700M hit; the rest extended theirs and flattered earnings
The enterprise layer is theater. 95% of GenAI pilots show no measurable P&L impact . Gartner: 40%+ of agentic projects cancelled by end-2027; ~130 genuinely agentic vendors out of thousands claiming the label . 75% of leaders claim adoption; 11% have production systems . 88% usage, 80% no EBIT impact.
Herd mentality, groupthink and mania-
Situational Awareness- in what world does a fired 23 year old Open AI employee publish a 165 paper arguing Artificial General Intelligence arriving by 2027 by way of some amorphous argument the machine will teach itself through sheer brute force of compute scale. This manifesto is then converted into a hedge fund where the fund lands millions and then billions of dollars to invest. This 23 year old then borrows billions to invest more, grows the fund by 439%, all to collapse on margin calls.
The talent mania (mid-2025). Meta offered $100M signing bonuses, packages reportedly up to $300M over four years, Zuckerberg personally cold-calling researchers for a 50-person superintelligence lab and then paying ~$15B for 49% of Scale AI substantially to acquire one 28-year-old CEO . The groupthink tell: roughly 1,000 people worldwide were simultaneously deemed to hold all the value, by every lab at once, meanwhile new-grad hiring at Big Tech collapsed to 7% of hires, down 50%+ from 2019 . An industry that claims AI will replace all human labor concluded that ~1,000 specific humans are irreplaceable. At the same time. With the same list of names.
The no-product lab mania (2025–26). Ilya Sutskever's Safe Superintelligence: $32B valuation in April 2025 with no product, no API, no published research, interviewees asked to surrender their phones to a Faraday cage and then Nvidia ploughed another $5B into it on July 27, 2026 . Mira Murati's Thinking Machines: a $2B seed at $12B; the largest seed round in tech history. At signing, "no specific product yet, let alone any revenue" . By late 2025 it was in talks at $50B, its lone shipped product was a fine-tuning API, and a co-founder had already walked to Meta
And lastly, because I just love Elon and his chainsaw-
The set-up: SpaceX didn't just allow retail into the IPO; it engineered it. A full 20% of the offering was allocated to mom-and-pop versus the typical 5–10%, and the crowd responded with over $70 billion in retail orders. Nasdaq rewrote its own rules to jam SpaceX into the Nasdaq-100 in fifteen days instead of the usual three months, turning every index fund and 401(k) in America into a forced buyer. One retail investor on Reddit said “The only people who actually 'lost' money are those who bought high and sold low.” This “rocket scientist” lost 50% of his investment blaming himself instead of the culprit that engineered the atmosphere.
If you have read this far you will see I have spread throughout the narrative irony and humor, but the stakes are real and potentially catastrophic. In every crisis, tragedy, economic downturn from the 1940’s though the 2008 financial crisis the failures of imagination, of exuberance, of group-think, herd mentality and mania; people died. Families were forever changed. In one tower on September 11, was an ex New Canaan High School Hockey player who I coached. He was killed. He was the one of nicest, hardest working and decent kids I have ever coached. That family still has not recovered. Shit, I haven’t recovered.
These failures cost lives, livelihoods, relationships, and multitudes of orders of consequence beyond. And, all of it, not some of it, ALL OF IT, could have been foreseen and acted upon, We now sit on the very same precipice and at another inflection point. Do we ignore the lessons of the past or do we IMAGINE our way towards better outcomes and a better future? Do we avoid tragedy, economic collapse, livelihoods lost and ruin or do we, for once, get our heads out of our asses and act?
