Goodbye, Programmers: AI's Token Economics Make Human Coding Skills Worthless
The programming profession's wage premium is structurally at risk: if AI can replicate top-tier coding work at token prices, salaries and software prices both collapse. Western developers who stay in pure engineering roles face a forced move into product definition or business management, or they compete against marginal electricity costs.
The claim that AI can reproduce John Carmack's year-long, 80-hour-a-week Doom effort in hours, and reverse-engineer Photoshop's core algorithms in weeks, anchors a wide-angry take on the collapse of programming as a career. Canva's November 2026 layoff of 22% of staff, over 1,200 people, is cited as an early sign that product and engineering workflows no longer need the same human headcount. The economic logic: when output is priced per token rather than per talent, human intellectual labor drops to electricity costs, and the technical value inside any software product approaches zero.
The future workforce, as described, will be a handful of product partners defining things and thousands of AI agents doing the build. Fewer than 15% of today's developers can make that shift, the piece argues. It also concedes that AI has not yet opened new consumer markets, and that the supply-side valuation (Nvidia, model vendors, token relay stations) may be a pure capital bubble if it fails to produce matching consumption. Long-term, AI's self-improvement turns the supply side into a capital-intensive industry and sets off a brutal price war.
The bottom line, in the writer's own words: human brain work is no longer worth anything, and the personal belief in intellectual value has collapsed.
The author's concession that AI has not created new consumer markets undercuts the central claim: if AI only eats existing markets, the collapse in software prices is a transfer of value, not a total devaluation of intelligence.
Replicating Doom in hours is a benchmark for raw code generation, but it doesn't compare to the design decisions and creative problem-solving involved in making a game; the essay conflates execution with creativity.
The bubble warning on the supply side implies that the token-pricing model may not be sustainable; if model vendors collapse, the cost advantage that makes AI cheap could evaporate.
Hiring thousands of AI agents still requires a handful of humans to define product requirements, so the 15% survival rate the author predicts assumes those humans are scarce and highly skilled, not just any programmer.
The essay's thesis splits opinion sharply. One detailed counter reads the AI boom as a product of compute overcapacity and investor optics, not real demand, while still conceding that AI flattens technical barriers and lets a novice replicate and undercut products. That view is countered by a blunt dismissal that the whole claim is AI wish-fulfillment (爽文), which itself gets challenged. The emotional range runs from gleeful liberation from exam hell to a classical lament over an inevitable collapse.
Actually, programmers are not that many; probably just over half. The emergence of AI short dramas can be simply understood as AI compute capacity being surplus (there isn't that much demand) needing to give investors a good-looking number; for most people it can be used as a replacement for search platforms, but if it requires payment, it probably won't happen. AI has indeed smoothed away technical barriers in most industries — for example, programming requires years of accumulated experience, but now a complete novice can use AI instead; a good idea and the rapid implementation and practice of a need no longer require waiting weeks or months. It can even reverse-engineer a good product and quickly copy it, then wear out the original through price wars. AI can completely let a 60-point person do work that requires someone above 85 points. Is programming still a special industry requiring advanced skills? Maybe not anymore, just like the current devaluation of degrees.
Every industry has a boom period; it's just that its time has passed.
AI wish-fulfillment fiction
Not really [I want some quiet]
The great building is about to fall, what can I do?