Artificial intelligence accelerates discovery of quantum circuits and molecules for new drugs
Artificial intelligence is already being used to design more efficient quantum circuits and identify promising molecules for future drugs. These advances help researchers explore complex pathways, although there remains a gap between a good computational response and the practical application of the technology.
A 53-hour search
In a recent experiment, Anthropic’s Claude model was tasked with creating quantum circuits—sequences of operations that guide calculations in these machines. For each proposal, an independent evaluator checked the result and assigned a score. The system used this feedback to refine the next attempt.
In seven out of eight challenges, the circuits found required fewer operations and fewer qubits, the basic units of quantum information. In the eighth case, it was possible to reduce the number of operations in exchange for a larger number of qubits.
One of the searches lasted 53 hours, with minimal human intervention, according to Anthropic. The result represents a saving in the design of calculations, but turning it into a practical advantage still depends on machines capable of executing them.
The race also goes through passwords
The prospect of more powerful quantum computers worries those who protect confidential information. Part of current cryptography relies on mathematical problems that are difficult for classical machines to solve, but that could be solved by sufficiently advanced quantum systems.
There is no exact date for this change. Still, delaying the update of data protection until it occurs would be risky. The NIST, the American Institute of Standards and Technology, warns that information can be collected now for an attempt to decipher it in the future.
In 2024, NIST published the first three final post-quantum cryptography standards. They were developed to resist this type of threat and work on conventional computers.
The research with Claude fits into this scenario of computational advances, without demonstrating a breach of current security. For companies, preparation involves monitoring the evolution of machines and updating information protection before this becomes an emergency.
From circuit to laboratory
In drug research, AI helps tackle another vast area of possibilities: interactions between molecules. AlphaFold 3, presented in 2024 by Google DeepMind and Isomorphic Labs, predicts structures involving proteins, DNA, RNA, and small molecules.
These predictions can help scientists investigate how a compound fits into a protein—a key clue for finding treatments. Knowing this interaction allows for formulating hypotheses about how to interfere with processes associated with diseases.
AI provides hypotheses to take to the laboratory, where another decisive stage of discovery begins. The model itself has limitations. According to EMBL-EBI, AlphaFold 3 produces static structures that do not capture the full dynamics of biological systems and can make errors in predictions.
Then, a drug candidate still needs to show how it behaves in the body. Clinical studies investigate doses, adverse effects, and benefits, as explained by the FDA, the U.S. regulatory agency.
What changes in the clock of science
AlphaFold does not depend on the progress made with Claude, nor on a quantum computer. They are examples of a broader change in research: using computational models to explore solutions and guide the next steps of work.
The benefit lies in reaching a hypothesis that deserves to be tested more quickly. Knowing how much this shortens the complete development of a technology requires monitoring the rest of the trajectory. A search can take hours; the answer to what it will allow us to build comes later.
Frequently Asked Questions
Did Claude create a quantum computer?
The model helped find more efficient circuits for specific tasks. The research deals with the design of calculations, not the construction of a processor.
Does AlphaFold need quantum computing?
No. Its molecular predictions are made by an artificial intelligence model that does not depend on quantum hardware.
Can AI create a drug on its own?
The examples presented show support for research. Confirming whether a drug candidate works and is safe requires experimental investigation and studies in people.








