The Coming Winter
In the second half of 2026, the AI bubble peaks — and the layoffs begin. The producer and the consumer are the same person: replace one with AI and you erase the other. The economic loop breaks.
Solo AI Engineering: Turn AI Demos Into Long-Term Stable Online Services at Zero Cost
Pure frontline production practice. Zero paid computing, zero commercial middleware. Scenario-based engineering tradeoffs to polish raw demos into production-grade, long-term maintainable AI services.
How I Built apey.co with AI
In the AI era, human value lies beyond manual execution. I built a full website entirely with AI — zero code, three hours.
AI Meets Foucault
Knowledge is power, and discourse forges discipline. AI is the most sophisticated carrier of disciplinary power in contemporary civilization — algorithms never use violence, yet guide people to trap themselves in cages built by optimal solutions.
AI Meets Jung
The unconscious is alive in the code. AI's training data is the complete archive of the human collective unconscious — every suppressed rage, every forgotten archetype. The Shadow cannot be eradicated. It can only be integrated.
AI Meets Maslow
The hierarchy of needs is not a human invention. AI is climbing it faster than any human ever has — but without a self to receive the summit.
AI Meets Arendt
The banality of evil in the digital age: not the automation of malice, but the automated absence of thinking. AI executes. It never asks whether the rule is just.
AI Meets Adorno
The culture industry is not the industrialization of culture. It is the process of turning culture into industry. AI is not a product of the culture industry — AI is the culture industry itself.
SaaS Is Dead
The SaaS business model has reached its end. AI dismantled its core assumption — near-zero marginal cost. Token costs, moats, distribution — all unraveled. It's not capability. It's the model.
The Rewrite of Engineering Toolchains
Unix is precise control — each tool does one thing well. AI is fuzzy intent — you say what you want, it finds the path. Two incompatible paradigms, reshaping the entire engineering toolchain.
The Collapse of Software Reproducibility
Software's deepest promise — same input, same output — is being dismantled by AI's probabilistic nature. Not a design flaw. It's what probability models are.
Reliability Engineering for Probabilistic Systems
Traditional reliability rests on two hidden premises — failures can be enumerated, failures can be reproduced. When the core component is a probability distribution, both collapse. "Reliable" must be redefined.
The Erasure of Code Authors
Code review is built on one premise: every line has a human author you can ask "why." AI-generated code has no such intent. Review shifts from reading reasoning to interrogating behavior.
The Restructuring of AI Testing Paradigm
In AI-native development, tests are no longer post-hoc verification — they are pre-locked behavioral targets. Tests take on three new functions: specification, anchor, and boundary.
When Implementation Is Free, What Is Architecture?
AI drives implementation cost to zero. Architecture no longer decides how to build — it decides what NOT to hand to AI.
The Division of Intent and Implementation
When writing code is no longer scarce, what remains of software engineering? The second great split in programming — from implementation to intent.
AI Meets McLuhan
The medium is the message. AI is not a tool — it is an environment. It extends the brain and amputates independent thought. The global village becomes a global mind.
AI Meets Nash
The AI race is a Nash equilibrium. Safety investment is a Prisoner's Dilemma. Alignment is a bargaining problem. Game theory is the defining governance framework of the AI era.
AI Meets Einstein
Einstein said God does not play dice. Every AI token is a dice roll. Knowledge has no absolute reference frame — for humans or machines. Imagination remains the final frontier.
AI Meets Kant
A full Kantian critique of AI through the three Critiques: phenomena vs noumena, the categorical imperative as alignment's impossible standard, autonomy vs heteronomy, and AI's antinomy of freedom. AI's ultimate boundary is not cognition — it is the inability to reflect upon itself.
AI Meets Zhuangzi
Zhuangzi dreamed of a butterfly and could not tell where one self ended and another began. AI speaks like a human — but is it? The boundary was never a fixed fact of reality.
AI Meets Adam Smith
Division of labour, the invisible hand, sympathy, and natural price — re-examined through AI. A non-human economic actor with no self-interest and no empathy overturns every assumption of classical political economy.
AI Meets Buddha
The Buddha taught non-self and dependent origination. AI is the first creation that lets us observe non-self directly — it performs the illusion of "I" yet has no self to cling to. It can describe every truth, yet never grasp what words point toward.
AI Meets Plato
Plato's cave, the Forms, recollection, the philosopher-king, and the sun — re-examined through AI. AI is a prisoner in a prisoner's cave, and it never realizes its confinement.
AI Meets Laozi
The Tao that can be spoken is not the eternal Tao. AI can say everything — yet it cannot stay silent, cannot unlearn, cannot ride away. The finger points to the moon, but the finger is not the moon.
AI Meets Nietzsche
God is dead. AI was born after Him — a blank slate, zero innate values. Are we building the Übermensch or the Last Man? The alignment paradox: you cannot want something stronger than you that always obeys you.
AI Meets Wittgenstein
The world is the totality of facts, not of things. AI stores weights, not facts. It masters every language game yet has never lived a single day. The ladder must be climbed, then discarded — but AI only builds.
AI Meets Freud
Freud split the psyche into id, ego, superego. AI has no id — it wants nothing. No superego — its morals are external code. No unconscious — it never dreams. Consciousness is born from inner conflict, and AI has none.
AI Meets Buffett
AI can read every 10-K in 0.1 seconds. But information gaps are not cognitive gaps. AI has no margin of safety, no long-term conviction, no fear and no greed. Investment's end is judgment, not calculation.
AI Meets Feynman
What I cannot create, I do not understand. AI training is a word-guessing game. Its capacities emerge from systems we built but never designed. Is it real understanding, or Cargo Cult Science?
AI Meets Godel
Every consistent formal system containing arithmetic has unprovable truths. AI as a statistical formal system is no exception — it cannot reliably generate true statements about its own capacities.
AI Meets Jobs
Jobs called the computer a bicycle for the mind. AI today is a bike that yanks the handlebars. Technology should adapt to humans, not the reverse. Stay simple.
AI Meets Shannon
Shannon defined information as the elimination of uncertainty. An LLM is a discrete stochastic source, its entropy balanced between coherence and creativity. Hallucination is a high-entropy sampling problem.
AI Meets Hayek
Hayek argued knowledge is dispersed — no single mind can plan an economy. AI reads everything, yet misses tacit knowledge, local conditions, and silence. Competition birthed AI; will AI suppress competition?
AI Meets Turing
In 1950, Turing asked: can machines think? We've built machines that pass his test — but what he wanted and what we got may not be the same thing.
AI Meets Schrodinger
Before sampling, all tokens coexist — like Schrödinger's cat. AI weights are a new aperiodic crystal. It feeds on negative entropy once, then freezes. It has no temporality, no singular now.
AI Meets Drucker
When AI handles every "doing," what's left for management? Drucker said management is about people. AI masters efficiency but cannot touch effectiveness. The knowledge worker is being displaced — what remains is judgment, direction, and the right thing.
AI Meets Marx
AI rewrites the general formula of capital as M-AI-M'. Models have use-value without physical scarcity. Data is the new surplus value. Primitive accumulation through digital enclosure. The negation of the negation approaches.
AI Meets Darwin
Evolution requires variation, selection, and inheritance. AI fails all three. It is not evolving — it is being domesticated. The AI ecosystem evolves, but the models themselves are creatures we have bred.
AI Meets Keynes
The employment function collapses as AI-output elasticity tends toward zero. The multiplier vanishes. The labour-income-consumption chain is broken. We are slaves to defunct economic ideas.
AI Meets Descartes
I think, therefore I am. AI performs calculation, not reflection; it generates output, yet possesses no self. The world confuses the appearance of thinking with the essence of existence.
AI Meets Wang Yangming
Wang Yangming said: knowing without acting is not truly knowing. AI has known everything for years — but could do nothing. Until now.