Artificial intelligence is a sharp break, not just another tech revolution
The end of history’ is a fool’s title. It invites mockery the way a tall poppy invites a scythe. When Francis Fukuyama used it in 1989, it was ridiculed. Critics pointed to wars and coups as proof that history was alive and well. They missed his point, but their reaction birthed a rule: never use a title that claims the past is over.
Let’s break that rule to make a narrower, uglier point. Historical analysis, treated as our primary compass, has become misleading. The temptation to look backward is a biological instinct; when the forest creates a new sound, the amygdala scans memory for a match. A twig snap may be a predator; a rustle mere wind. History works as a guide most of the time, which is exactly what makes it lethal when it does not.
In AI, since the release of ChatGPT and DeepSeek’s open-weight shock to the rise of autonomous coding agents like Claude Code, breath-takers have become routine. We no longer debate if productivity will rise; we ask if our notions of a ‘job’ and a ‘firm’ were ever stable. Pattern-matching still works for the cyclical theatre of stock markets, but in the real world, reaching for historical reassurance is risky.
Tyranny of the decimal: There is a mathematical violence in recent developments that analysts miss. For years, the consensus was that AI could assist with code but not produce reliable systems because even a 1% error rate compounds brutally. Statistically, a 1% ‘hallucination’ rate per line guarantees the failure of a 500-line program. But we have moved from GPT-4 to a ‘collapse of the decimal.’ When the error rate drops to 0.01%, possibilities explode.
This is a phase shift. Systems are now re-factoring 10,000-line repositories and joining their own improvement loops. This is recursion—a cycle that changes its own geometry. It is the ‘Super-Moore’ era, where capability grows fast to render history irrelevant.
Dichotomy of time: We are trapped between decades and Wednesdays. Secular forces like labour restructuring and the commoditization of intelligence are tectonic plates that move over decades. But our grasp of possibilities is measured in hours. A conviction held on Monday can be obsolete by Wednesday. Our intuitions of exponentiality are weak.
Consider the paper-fold: if you fold a sheet of paper 42 times, it doesn’t become a thick notebook, it reaches the moon. For the first 30 folds, progress looks linear. Then it goes lunar. We are currently around fold 35 in AI.
The aesthetic of authority: A retreat to history is an aesthetic choice. A chart stretching across decades, with regression lines bending toward a reassuring mean, has an aura of inevitability. It acts as a neurological sedative. It suggests that reality remains tame.
By contrast, projecting forward from current observables feels reckless. But logical extrapolation compounds brutally. Our incentive is obvious: history holds credibility, but logic vulnerability. This is why analysts alter assumptions but do not rethink them, often folding world-altering changes into old narratives.
Today, understanding competitive advantage requires confronting abstractions like ‘CoWoS’ semiconductor packaging or recursive agentic workflows. These cannot be reduced to a viral tweet. The latest developments in autonomous reasoning are potentially more significant than the arrival of ChatGPT itself. Yet, they are almost impossible to discuss without their sounding like a ‘castle of assumptions.’
The ‘as we know it’ filter: We typically append ‘as we know it’ to phrases like the ‘death of software.’ It was not a hedge but a diagnostic tool. Industries rarely vanish overnight; as with car-making in the US, they ‘Detroit-ify.’ They lose their centrality and pricing power while smart people within them make Herculean efforts to re-pivot.
Many of these pivots will fail, rendered obsolete by a ‘Wednesday shock’ that arrives before a new strategy gets a chance. Others may succeed for a season, only to realize the ‘new normal’ was just another fold on the way to the moon. Acknowledging this instability is the only path to agency. The old map is gone; the only mistake greater than having no map is insisting on using one from the previous century.
Honesty in front of a mirror: In private, we do not get to outsource honesty. The most dangerous phrase in analysis is not ‘this time is different,’ but we cannot know’ when used as an excuse to stop thinking. The humility we need now is of the real kind. We must admit that our experience, which served well for decades, may now be pointing us in the wrong direction. The cost of that error is not abstract. It is capital trapped in dying assumptions and a generation of students trained for jobs that machines will perform more effectively before they graduate.
History has not become useless, but its reliability has become uneven. Pattern recognition, part of our survival instinct, is beginning to hallucinate continuity where none exists. The past does not announce its lost jurisdiction; it continues to offer answers even after the questions have changed.
AI will not wait. We must spend less time debating ‘hype versus reality’ and more time observing. We build monuments to the mean, but we are buried in the tail-ends of the bell curve; and these tails could soon form the bulk of it.
The author is a Singapore-based innovation investor for GenInnov Pte Ltd.
Post Comment