The Main Shift in the Labor Market: The Future of Professions in the AI Era

Artificial intelligence does not take your profession — it takes the entry into it. People were told to learn programming, and now the market says AI will write the code, so an illusion arises that everyone will become a programmer. The host calls that untrue: generating code does not mean creating software, and assembling an app by vibe coding does not mean building a system for a real company with customers and money. AI pulls the simple tasks out of a profession, and the winners will be those who understand business, product, data and processes and build AI native systems.

As of early June 2026 AI has not fired everyone, but the market already looks like the story of a break in the entry into professions. The blow falls not on the programmer's profession but on junior roles, routine tasks and entry-level employees — administrative work, part of analytics, template content, support. The host gives a case: he asked a director to find data and an analyst for a report, but his own request to plain ChatGPT had already produced the result, and he decided to let those people go. Entering any business, he now asks for the raw data and runs it through AI models.

Vibe coding, the host says, is still working with code: the term was introduced by Karpathy and meant that Claude Code or ChatGPT Codex help write code. For Alexander Volchek, Codex is not about vibe coding but about developing AI native systems: software developed purely by AI. He never opens the code, and when Codex or ChatGPT offers to run a check or a launch in a database, on a server or in GitHub, he answers: "I won't do that, you have to do it yourself." Meanwhile the difference between ChatGPT 5 and 5 Pro, or between 4.8, 4.8 Max and 4.7, is understood by perhaps one in fifty.

The host likes the position of NVIDIA chief Jensen Huang: people talk about AI layoffs, while in reality they are tied to companies' financial results — AI is merely the justification. By the Stanford AI Index 2026, layoffs are not yet mass, but the market for junior specialists is shrinking. Who are juniors? The world's best AI startups, worth billions, were built by people around twenty — yesterday's junior specialists who created something. Hence his advice not to pay for nominal experience: an accountant with six months' experience is often more professional than one with twenty years.

By the Stanford AI Index 2026, one of the most important studies on the subject, generative AI reached 53% adoption in three years — faster than the PC or the internet. OpenAI has about a billion weekly users, Gemini's monthly users are approaching a billion, and Google AI Mode has reached a billion. But the effect on labor is uneven: among developers aged 20–25 employment has fallen by almost 20% in two years. Such studies, the host adds, go stale and are made on their own samples: companies in Belarus and Russia should not measure themselves against American or Indian ones.

Millions of dollars invested in AI have not increased corporate profits, and many companies do not want to fire anyone to deploy it but to strengthen efficiency. The host himself is interested not in layoffs but in adapting employees and partners. Understanding AI means working at it constantly with paid subscriptions and creating things for yourself. One example is his sister's idea of exporting all her ChatGPT chats into Codex to assemble "cores" of knowledge, which are then attached to a ChatGPT project. The opposite example is Zoom's cheap-model AI summary, which distorted his session.

A profession is not a single object, it breaks down into tasks: AI first takes pieces of work, then changes team structure and hiring, and only then come layoffs; the economy, law and education react even later. The host has four children. His daughter is learning to drive at fifteen, but his three other children hardly will: in a year or two the Tesla Cybercab will appear, with no steering wheel and no license needed. In Belarus, Russia or Italy that will not come soon, in China and the US it will, and how to decide in your own country is the main question.

To his nephew, a polytechnic student in Poland, the host advises creating something every day in the most fashionable AI systems, and for that subscribing to ten or twenty well-known people — Karpathy, Dario Amodei, people from DeepMind, Anthropic and OpenAI, xAI engineers and Chinese engineers — and trying to reproduce the cases they post. That is the only way to stay in the top layer. Lectures are unlikely to help, and tasks handed down by a business are one of the big deceptions: eight out of ten businesses say they use AI, but in the details it is clear how little people understand it.

People do not understand the difference between the $20 and $200 plans, and heads of American corporations say a programmer earning more than $500,000 a year who does not eat that much in tokens is of no use to them. To the objection about a $300 salary he answers with his nephew with a cheap Claude subscription: buy Opus Max for $200 — it is what will let you earn tens of thousands or millions. He promises no quick money: it is about knowledge for a lifetime, and an AI subscription cannot be treated like Netflix or a restaurant, although even his partners do not buy the pro version.

Google and Meta put up cameras to see what employees do; the host would do the same for efficiency, not to replace people: optimizing manual labor by 90–95% is for x2–x3 in the business, not to save 10% of payroll. Microsoft, analyzing 200,000 Bing Copilot conversations, found that AI is most often used for gathering information, writing and learning, but office workers will not disappear. A new class of people will win — those who understand business and customers and who used to be constrained by staff, analysts and programmers; marketers without strong analytics will not succeed.

Contrary to claims of two or three years ago, plumbers' and electricians' labor will, the host believes, increasingly be destroyed: ordinary people with AI will do that work themselves and specialized robots will keep improving, builders included. Where work was hard to automate, the market will be restructured phenomenally. Elon Musk's driverless $200-a-month Cybercab is a fundamental revolution: such a car will take the children to school and rebuild daily life, though the laws push it beyond a two-to-four-year horizon. Tesla, meanwhile, is focused on robots and wants to produce a billion.

About two years ago Elon Musk put the probability that AGI becomes the dark side of the world and takes everything over at 20%. In the clip, he says something smarter than any human may appear within a couple of years, and something smarter than all humans combined by 2029–2030; a good outcome he puts at 80%. The host allows both a severe crisis with work and the opposite — AI letting people live better and work less, although talk of state payouts was a 2023–2024 trend, not a 2026 one. The main thing is to allow that in twenty years one's profession and earnings may be rebuilt completely.