"AI is better than me": the vertigo of a researcher facing Claude and ChatGPT

The dizzying observation of a French researcher

Antoine Tilloy, a theoretical physicist at CNRS, has published a thread that has caused a stir within the scientific community. He explains that artificial intelligence now surpasses humans in almost all intellectual fields. According to him, theoretical research will "profoundly change with the exceptional progress of new artificial intelligences."

The research in theoretical fields will profoundly change with the exceptional progress of new artificial intelligences, and I believe we all wonder what will remain...
--- Antoine Tilloy (@AntoineTilloy) August 10, 2026

The researcher mainly describes an extremely rapid evolution of the most efficient models:

It’s as if I were talking to an excellent colleague, brighter than me in almost everything, and who executes about 1,000 times faster.
Antoine Tilloy, theoretical physicist at CNRS

Tilloy believes he retains an advantage on certain problems he has studied for several years, especially when it comes to taking a step back.

But this boundary seems to be shifting: AI no longer just executes its instructions and begins to suggest the next steps of its research.

More and more, Fable, after executing what I asked, proposes the next steps, which are quite often the right ones.
Antoine Tilloy, theoretical physicist at CNRS

The researcher even calls for "seriously thinking about a future, potentially quite close, where AI dominates us in absolutely all intellectual fields."

However, he believes that the current progression "resembles as much a sigmoid as an exponential": capabilities could continue to explode or eventually reach a plateau.

Tilloy also indicates that AI has only been truly "unavoidable in theoretical physics for a maximum of one year." He personally uses subscriptions to Claude Max and ChatGPT, while CNRS rules favor Mistral AI models for professional use.

Claude advances a 167-year-old mathematical problem

This testimony comes as an unpublished research version of Claude, the Anthropic model, has just made progress on a problem related to the Riemann hypothesis, one of the great enigmas in mathematics since 1859.

Claude has not solved the hypothesis. However, the model has raised the known lower bound of the proportion of zeros of the zeta function located on the "critical line" from 41.6% to 67.2%.

To achieve this result, Claude first explored about 650 unsuccessful leads. The system then coordinated around sixty sub-agents for a day and a half, performed thousands of numerical checks, and consulted 54 scientific articles to verify that its result had not already been discovered.

We asked an unreleased research version of Claude to take a stab at the Riemann hypothesis.

It didn't solve it, but it did make strides on a related problem: it increased the lower bound for the fraction of zeros of the Riemann zeta function that satisfy the hypothesis from...
--- Anthropic (@AnthropicAI) August 10, 2026

The result has been studied by mathematicians at Anthropic and examined by specialists Brian Conrey and Dan Goldston. Claude has also produced a formalization of its proof with Lean, which successfully passed its verification tool. However, a broader review by the scientific community is still necessary to definitively establish the significance of this advance.

What role will remain for researchers?

These spectacular results recorded by Claude show that AI is making rapid progress and can help solve enigmas that scientists have sometimes struggled with for centuries. For Tilloy, this evolution could profoundly change the skills valued in research.

Pure executors, reliable and rigorous, who were previously valued in collaborations, are losing much of their added value.
Antoine Tilloy, theoretical physicist at CNRS

If AI becomes extremely efficient in execution, the value of the researcher could shift towards problem formulation, creativity, and the ability to maintain a holistic view.

A boundary on which humans still retain an advantage, according to Tilloy. In theoretical physics, Fable sometimes remains "less good than [its] best colleagues" when it comes to inventing a new mechanism or a new idea. "I don’t know if that will remain the case," he acknowledges.

For Tilloy, the shift is already underway, even if the models were to stop progressing today.

Even if the models stopped progressing today, the field would be profoundly changed, probably as much as with the advent of computers.
Antoine Tilloy, theoretical physicist at CNRS

The physicist ultimately describes himself as "Promethean" in the face of the eventual arrival of a superintelligence: if it is achievable and can be made safe, he believes we should seek to develop it. But he also acknowledges that he hopes this goal remains "intrinsically difficult," so that there are still many years of human research ahead before achieving it.

This content is provided for general informational purposes only and doesn't constitute financial, investment, legal, or tax advice. Any events, rewards, online promotions, or related information mentioned herein should not be considered a recommendation, solicitation, or invitation to purchase, sell, trade, or otherwise deal in any crypto assets. Crypto assets are highly volatile and may result in loss. The availability of WEEX services, products, and related events may vary by region. You are responsible for ensuring that your participation is in accordance with applicable local laws and regulations.