Should we be worried about the fate of the world: three reasons to ignore the Citrini report

The Citrini Report is a super-optimistic report. A close look at this text shows that its authors believe in the omnipotence of artificial intelligence much more than all the AI analysts in the Trump administration: if the Trump administration were so confident in the omnipotence of AI, they would only think about regulating it, according to AI researcher Mikael Gorsky.

The report is written as a report from the future, I'll note that I found it interesting.

- AI-agent coding tools such as Claude Code, Codex, Cursor, Lovable and others have made a quantum leap in late 2025. A single competent developer working with Claude Code or Codex is capable of replicating the core functionality of a mid-market SaaS product in a matter of weeks. It's safe to say that the implementation won't be perfect, but it will be good enough to make an IT director think: Why are we paying half a million dollars a year for a subscription?

- By early 2027, AI agents will be able to make consumer decisions autonomously, running in the background on phones and laptops. People who have never heard the word "agent" will use them in the same way they used to - for autosuggestion or spell-checking. Commerce ceases to be a series of individual human decisions and becomes a continuous optimization process.

- By March 2027, the average American will consume 400,000 tokens per day - ten times more than at the end of 2026. Distilled (compressed) LLMs will work on phones and laptops, not just in the cloud, radically reducing the cost of usage.

- Subscriptions that automatically renewed for months without usage. Trial rates that quietly doubled after the end of the trial period. Insurance policies where the entire renewal model was based on customer laziness. Realtor commissions that stayed at 5-6% for decades thanks to information asymmetry. All of this will come crashing down when decisions are made by autonomous agents who aren't lazy, take their time and remember to check alternatives.

- AI agents will find the cheapest route for each payment. Card fees of 2-3% become an obvious target for optimization. Agents will move to stablecoins via Solana or Ethereum L2, where settlement is almost instantaneous and transaction costs are measured in fractions of a cent. Mastercard will record a slowdown in purchase volume growth in the first quarter of 2027.

- The catalyst for the retail sunset will be an open-source shopping agent from Alibaba. Within weeks, every major AI assistant will integrate an agent commerce feature. Consumers will easily adopt the new technology and AI-assisted shopping will become the norm rather than an exotic techie pastime.

- AI will completely replace customer support services. The familiar elements of customer support - tickets, their routing, and operator interaction management - will give way to AI systems that solve problems without creating tickets at all.

- AI agents will autonomously perform research and development tasks that used to take a human weeks to complete. These are no longer autocomplete code or chat assistants. These are systems that will take a task, decompose it, execute each step, validate the results, and deliver a finished product. Exponential growth will sweep away all notions of what is possible.

- AI will write virtually all of the code. The role of the human developer will shift from writing code to top-level coordination and flavor control. Coding as a craft requiring years of training and high pay will cease to exist in its former form.

- The best AI systems will become significantly smarter than almost all humans at almost everything. It won't be narrow specialists who are good at chess but don't understand humor. It's general intelligence that will surpass most humans in analysis, decision making, creativity, and coordination. And it will continue to get cheaper.

- Companies threatened by AI will be its most aggressive adopters. This will be different from the historical model, where established companies resisted new technologies and slowly died, like Kodak or Blockbuster. It will become impossible to resist - the stock will drop 40-60%, the board of directors will demand answers, and the only answer is to cut people and invest the savings in AI.

- Self-development will become a viable alternative to buying SaaS at the enterprise level. In-house teams will build prototypes that replicate the functionality of six-figure contracts in weeks. Even if self-development isn't always the best choice, the very fact that it's possible will change the negotiating dynamic. SaaS vendors will no longer be able to dictate terms.

- Fired programmers won't be able to move into "AI management" because AI will have already learned how to do that too. Previous technological leaps in human history have destroyed jobs and created others. But then each previous new job required a human to perform it. AI will improve in exactly those tasks that humans could be retrained to do.

- Labor's share of GDP will fall from 56% to 46% in four years, the sharpest decline in recorded history. Before that, the decline from 64% to 56% took four decades and was driven by globalization, automation, and the gradual weakening of workers' bargaining power. AI will make a similar drop in four years.

The authors of the report surpassed me in their super-optimism. The development of LLM is no longer just a technological advancement, it is the discovery of an unearthly intelligence. GPUs, when assembled in a big pack, acquire the ability to reason and think. But this, in my opinion, pales in comparison to what is written in the report.

First, in defining what provides companies with stability - replacing Doordash / Wolt / Uber.Delivery with wybcoding is unrealistic if you realize that the success of these companies is not in a nice interface, but in trusting relationships with the masses of couriers and restaurants.

And secondly, in predicting a world in which beings devoid of bodily experience infiltrate all areas of human endeavor.

But in general I completely agree with Citrini that AI can do anything and will defeat everyone in the world. More precisely - people armed with it will defeat everyone else.

What is harder for me to understand is the conclusion of this report - since AI can do everything, only AI will work, and solvent demand will remain only for what AI needs. Well, the disappearance of effective demand for "human" things will kill the world economy.

In essence, the thesis of the Citrini report boils down to the fact that if at the onset of radical economic changes there are no changes in the fundamental economic mechanisms and activities of states and government agencies, everything is lost. In such cases, Americans say "Duh", and in Russian-language Facebook, the author of the thesis "if you do not react to changes, it will be bad" will be called "Captain Obvious".

Except that in the real history of the real world we observe otherwise: all revolutionary changes are accompanied by changes in social and economic attitudes.

The first example is the industrial revolution in nineteenth-century England. The massive introduction of machine production destroyed artisanal labor and created monstrous living conditions in the cities. The first decades were exactly the nightmare that Citrini describes: rising productivity, falling worker incomes, concentration of wealth among factory owners. But then the economic mechanisms changed. Factory legislation emerged that limited working hours and child labor. Labor unions emerged and gained legal status. The state took over education, sanitation, regulation of working conditions. The tax system evolved from excise and customs duties to an income tax, allowing the benefits of productivity growth to be redistributed. None of these changes were inevitable or rapid. But they happened precisely because radical economic change created pressures to which society eventually responded by changing the rules.

A second example is the Great Depression and the New Deal in the United States. By 1933, unemployment had reached 25%, the banking system had collapsed, and farmers were losing land. The existing economic mechanisms - the gold standard, minimal government intervention in the economy, lack of deposit insurance and pension system - could not cope with the crisis. So they were replaced. Roosevelt created the FDIC, a corporation to insure deposits, the SEC, a commission to regulate securities markets, and Social Security for pensions. The Glass-Steagall Act separated commercial and investment banks. The state became employer of last resort for a time through the state WPA and CCC employment programs. The tax system was radically overhauled: the top income tax rate rose to 79%. None of these tools existed before the crisis. They were all created in response to pressures that the old mechanisms could not cope with. This is what makes Citrini's thesis trivial: the authors describe a world in which a revolution on the scale of the industrial revolution is taking place, but states somehow respond with the tools of 2025.

This article was AI-translated and verified by a human editor