An artificial intelligence system has solved one of quantum computing's most stubborn operational challenges: keeping the delicate laser systems that control qubits stable without requiring on-site experts. QuEra Computing announced that Anthropic's Claude AI agent successfully wrote and validated control software for laser systems in its quantum computers, recovering from faults in seconds compared to the five to ten minutes required by human specialists.

Why Quantum Computers Need Constant Babysitting?

Quantum computers are extraordinarily finicky machines. They rely on lasers tuned to precise frequencies to control atomic qubits, the quantum equivalent of classical computing bits. These lasers drift over time, and when they drift far enough, the entire system stops working until someone with deep expertise brings it back online. For years, this meant that scaling quantum computers required an army of highly specialized technicians stationed at customer sites, ready to respond at any hour.

QuEra's 256-qubit Aquila system, available on Amazon Braket, already achieves uptime above 99 percent through automated recovery from common laser disturbances. But the more complex laser failures have resisted automation because they require human judgment rather than following a fixed sequence of steps. A team of four QuEra specialists spent two to three weeks writing a recovery script by hand, and it only handled the specific failures they anticipated.

How AI Outlearned the Experts?

QuEra partnered with Anthropic to test Claude through the Model Hardware Standard (MHS), a new framework designed to let AI agents safely operate physical equipment in scientific research and advanced manufacturing. The MHS started as a collaboration between Anthropic and HHMI Janelia Research Campus and is currently in limited research preview.

The key difference: instead of writing a script once, Claude ran continuous experiments on a dedicated testbed, proposing fixes, testing them, reading results, and refining its approach overnight. Engineers set the boundaries and reviewed every step, but the AI covered hundreds of failure cases that no human team could work through manually.

The results were striking. Claude's software recovered the laser system reliably in 695 of 700 timed trials across seven different fault types. Most faults cleared in under six seconds, with the hardest cases resolved in roughly 10 to 14 seconds. When asked to improve the quality of the lock rather than just recover it, the AI cut residual noise by a factor of five and prevented the system from dropping out during unattended runs.

"For years the hardest part of scaling quantum computers wasn't the physics, it was the people driving at 2 am to fix a laser lock. We built a solution using the Model Hardware Standard to fix that: the lock recovers itself in seconds, verified every time, catching noise that's easy to miss by hand. We're building quantum computers that fix themselves," said Sergio H. Cantu, Vice President of Quantum Systems at QuEra Computing.

Sergio H. Cantu, Vice President of Quantum Systems, QuEra Computing

Perhaps most impressively, when pointed at a second laser wavelength, Claude worked out the settings from scratch in one unattended overnight run, a job that normally takes weeks of hands-on commissioning.

What This Means for Quantum Computing at Scale?

The implications are profound. Prior to this pilot, every laser in a quantum computer carried a standing claim on scarce human expert time. On-site recovery was required at any hour and could take up to half an hour per tuning session. Commissioning a new operating point took weeks. With AI automation, that burden essentially disappears.

For high-performance computing centers or national laboratories installing quantum systems on-premises, this is the difference between needing a resident laser specialist and needing very little of their time. As quantum computers add more lasers with each generation and deployments spread further from the engineers who built them, manual expert labor was becoming a hard limit on how many systems could be deployed and properly supported.

"We are among the best in the world at developing and operating quantum computers, and even for us, the cost of keeping these machines at peak performance is high. A customer expects the entire computer, and thus every subsystem, to hold itself together without a specialist in the room. This is why the results from the MHS research preview and Anthropic's frontier AI models are so meaningful. We are making it far easier and cheaper to keep our computers running at their best," stated Takuya Kitagawa, President of QuEra.

Takuya Kitagawa, President, QuEra Computing

How to Understand the Safety Framework Behind AI-Controlled Quantum Hardware?

One concern with AI controlling physical equipment is safety. The Model Hardware Standard addresses this by building safety into the framework itself rather than relying on the AI to make safe decisions at runtime. Here's how the approach works:

• Declared Boundaries: Devices declare their operating bounds, interlocks, and emergency stops in the standard itself, and AI agents inherit and operate within them by default.
• Human Oversight: Engineers set the scope of the AI's work, reviewed every step of Claude's solution, and decided what counted as proof of success before deployment.
• Inspectable Code: The software Claude produced is a conventional, fully inspectable program, not a neural network making opaque decisions at runtime, making it auditable and trustworthy.
• Real-World Testing: The testbed sat in a working lab with ordinary foot traffic and real environmental sources of error, proving the system works under actual conditions, not just controlled tests.

QuEra expects the same AI automation approach to apply to other subsystems in its quantum computers, and the pilot leaves behind the safety practices and measurements that enabled the second campaign to take one night where the first took weeks.

What About Quantum Computing's Other Challenges?

While QuEra tackles the operational side of quantum computing, other researchers are making progress on the algorithmic front. A team of Singapore-based researchers from the Agency for Science, Technology and Research (A*STAR), the National University of Singapore, and Nanyang Technological University demonstrated a hybrid quantum-classical method for molecular docking on an IBM quantum processor.

The researchers used a technique called full-basis encoding to represent more information with fewer qubits. In traditional quantum optimization, each choice in a problem gets assigned to one qubit. An 18-variable docking problem would normally require 18 qubits. By using three measurable properties of each qubit to represent three separate choices, the team represented an 18-variable problem with just six qubits and a 14-variable problem with five qubits.

For both molecular examples tested, the quantum hardware identified the same groups of compatible molecular contacts as classical calculations. The full-basis method found the best solution in about 74 percent of 1,000 runs for the 18-variable problem. While these results don't yet demonstrate quantum advantage over classical computers, they show how researchers might fit parts of drug-discovery problems onto current quantum hardware, which are limited by noise, errors, and the number of usable qubits.

Together, these developments show quantum computing moving from theoretical promise toward practical deployment. The challenge isn't just building better quantum processors; it's making them reliable enough that customers can operate them without a physics PhD on staff and flexible enough to solve real-world problems like drug discovery. AI is proving to be a crucial tool for both challenges.