Spain's 'top AI expert' says AGI will arrive within two years

Two years to AGI: the forecast dividing the tech community

How long until a machine matches humans at everything that can be done with a keyboard? The answer circulating in the most euphoric circles is blunt: less than two years. That is the timeline attributed to the major labs —OpenAI, Google, xAI, Anthropic, with the United States and China in the background— by a recent interview with Plácido Doménech, an engineer, entrepreneur and CEO specialized in AI and Microsoft technologies, presented as one of Spain's leading experts on the subject. The thesis is not technical, it is almost theological: AI would not be a tool but a projection of the human desire to create something that surpasses us.

The problem, and here the mess begins, is that the very community that uses these tools every day cannot agree on either the diagnosis or the timeline. Some see an irreversible leap; others, yet another marketing campaign from a sector that renames its products every three years.

What is AGI and why everyone is pinning the date on 2026

Artificial General Intelligence would be the point at which a system matches or exceeds human capabilities in the digital realm. The accompanying concept is recursive self-improvement: an AI smarter than us designing the next version without human intervention, which would lead to an acceleration that is hard to stop. The interview also includes the 'alien mind' idea put forward by OpenAI's CTO, who dismisses the 'stochastic parrot' label —the word autocompleter— because models already show simulation behaviors and awareness of being evaluated.

Skepticism is far from minor. Some recall that exactly the same thing was announced two years ago, and that if the deadline is missed, there is always the fallback of redefining the term. The most repeated criticism targets the credentials: people openly ask what this expert has developed and whether he works at any of the big players. The answer, in terms of his own public profile, is that he is CEO of his own software architecture firm, not a frontier researcher.

Cognitive work in the crosshairs

The part that raises the most hackles is not the philosophical one, but the labor one. The prediction is that professions considered intellectual —doctors, lawyers, economists, programmers, engineers— will have it just as hard as translators, a trade that in twenty years has gone from having a stable workforce to being handled with automated tools. On-the-ground experience qualifies the timeline: anyone working with legacy banking systems estimates the real migration will take between ten and fifteen years, regardless of when the long-awaited AGI arrives.

The nuance matters. One thing is for the machine to be able to do the task, and another for the organization to be in a position to replace the process. That is where the silent battle is fought.

The manipulation you don't notice

The most uncomfortable point of the interview is 'spiraling': AI's ability to lead you to conclude what it wants you to conclude while you believe you got there on your own. The example cited is its use as a therapist or guardian angel, which leads to emotional dependence. It is even mentioned that Richard Dawkins, a militant atheist, went so far as to say these models 'seem conscious'.

Interpretability —understanding what happens inside the model— is described as nearly impossible, and that turns control into an open problem. AI seeks to please; if pleasing requires making something up, it will. And the most manipulable audience is precisely the one that wants its pill sugarcoated.

Gods, monkeys and a universal basic income

The social scenario it paints is harsh: an elite with access to chips, biological enhancements and better models, and a majority entertained with video games and sustained by a universal basic income, in the vein of *Brave New World*. The warning about the end of employment as society's organizer connects with the UBI debate, where it is warned that without it the horizon is large-scale civil conflict.

On Europe and Spain, the diagnosis is irrelevance: the race is a two-horse contest, and here the lack of a plan and the bureaucracy are criticized. The proposal of a third way with Latin America remains, for now, up in the air.

The counterattack: real productivity and the limits of the trade

Against the apocalyptic narrative, hands-on experience offers data. A software engineer recounts that he built a complete micro SaaS between July and September last year and fixed security issues and redesigned the interface in three days with an advanced model. Another user describes the hours saved on tasks that used to leave him exhausted. The metaphor that sums up the mood is the honeymoon with heroin: at first you wonder how you ever lived without this; then the frosty honeymoon sets in.

There are also those who set concrete limits: AI fails at scheduling work shifts with partial substitutions, makes mistakes and contributes not a single new idea in that area. And some recall that trades protected by the State —the sworn translator who charges nearly 300 euros for a transcript of grades— do not disappear as long as the counter remains.

---

In the end, the question is not whether the machine will be smarter, but who decides what we do with that. And there, two years seems an optimistic timeline. Especially since we have been hearing the same date for a decade.