Insider Brief
- QuEra reported that Anthropic’s Claude developed software that can automatically restore a critical quantum computer laser system in seconds.
- The controller recovered the system in 695 of 700 timed trials across seven fault types and never falsely reported a successful recovery.
- QuEra said the approach could reduce the need for on-site specialists as quantum computers become larger and are deployed at customer facilities.
QuEra Computing reported that an artificial-intelligence agent developed software that can restore a critical quantum computer laser system in seconds, potentially reducing the need for specialists to maintain machines at customer sites.
The Boston-based quantum computing company reported in a news release that Anthropic’s Claude wrote and tested a control program that recovered a laser system from hundreds of simulated and real-world disturbances. The work was conducted through a research preview of the Model Hardware Standard, or MHS, a framework designed to let AI agents operate laboratory and manufacturing equipment within preset safety limits.
The results, also discussed on an Anthropic blog post, address a practical problem facing the quantum computing industry. As quantum computers move from research laboratories into national labs, supercomputing centers and other customer facilities, manufacturers need the systems to operate without constant support from the scientists and engineers who built them.
QuEra’s computers use tightly controlled lasers to manipulate neutral atoms, which serve as quantum bits, or qubits. The lasers must remain at precise frequencies. Environmental changes and other disturbances can cause them to drift, forcing operators to restore what is known as a laser lock before computation can continue.
QuEra has previously automated recovery from routine laser disruptions, according to the news release. The company said Aquila, its 256-qubit quantum computer available through Amazon Braket, maintains uptime of more than 99%. More complex failures, however, have continued to require judgment from specialists.
AI Develops Recovery Software
According to QuEra, four specialists previously spent two to three weeks writing a recovery script by hand. That program could address only the failure modes that the engineers had anticipated and included in the code.
The company then assigned the problem to Claude through MHS. The standard allowed the AI agent to conduct experiments on a dedicated laser testbed, evaluate the results and revise its approach. QuEra said the agent ran this process continuously, including overnight, and examined hundreds of potential failure cases.
Engineers determined the agent’s operating limits, reviewed its work and established the conditions required to demonstrate success. MHS also allowed the equipment to specify safety boundaries, interlocks and emergency-stop procedures that the agent was required to follow.
Claude did not remain in control of the laser after development was complete. Instead, it produced a conventional software program that engineers could inspect and validate. That program now handles recovery without an AI model making decisions while the system is operating.
In 700 timed trials covering seven types of faults, the resulting controller returned the laser system to its target state 695 times, QuEra said. It did not falsely report a successful recovery during any trial. The company attributed the five unsuccessful attempts to a condition in the experimental apparatus rather than a defect in the controller.
Most faults were resolved in less than six seconds, while the most difficult took about 10 to 14 seconds. QuEra said a specialist would typically require five to 10 minutes to resolve comparable problems.
The testbed operated inside a working laboratory, exposing it to ordinary foot traffic and environmental disturbances rather than only controlled test conditions. QuEra said the controller recovered from every naturally occurring disruption during the pilot without human assistance.
The AI agent was also asked to improve the stability of the laser lock. QuEra reported that its settings reduced remaining noise by a factor of five and prevented the system from losing its lock during unattended operation. The company said an independent measuring instrument confirmed that the results matched an experienced specialist’s manual tuning while correcting a problem the manual process had missed.
QuEra then applied the approach to a laser operating at a second wavelength. The agent determined the settings during one unattended overnight run, according to the company. QuEra said comparable commissioning work would ordinarily require weeks of hands-on effort.
Reducing the Need for On-Site Specialists
Laser maintenance could become a larger obstacle as neutral-atom quantum computers grow. Newer machines use more lasers, while installations at customer facilities may be far from the engineers with the necessary expertise. QuEra said manual tuning can require up to 30 minutes in some circumstances and may demand on-site assistance at any hour.
“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.
The pilot involved one subsystem and does not demonstrate autonomous maintenance of an entire quantum computer. Quantum systems contain many other components that require calibration, monitoring and repair. QuEra said it expects the same development method could be applied to additional subsystems, although the company did not provide test results for those applications.
MHS began as a collaboration between Anthropic and the Howard Hughes Medical Institute’s Janelia Research Campus. It remains in a limited research preview while its safety design is evaluated. Access is currently available by application.
“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,” said Takuya Kitagawa, president of QuEra. “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.”
QuEra is working with Amazon Web Services on cloud delivery of its planned Libra system through Amazon Braket in 2028. It also has partnerships with Hewlett Packard Enterprise for integration with on-premises high-performance computing systems and NVIDIA for accelerated computing.
