The Lyceum: Quantum Intelligence — Jul 23, 2026

Infrastructure, not breakthroughs, defined this week. China stretched stored entanglement across an intercity fiber link, the Internet Engineering Task Force pushed post-quantum cryptography toward deployable protocol language, and Washington added money and budget machinery for fault-tolerant computing. The field’s center of gravity is moving from can this work? to the less glamorous question that decides industries: can anyone build, standardize, and operate it reliably?

Long-distance entangled photons are difficult. Entangling devices that can store quantum information is harder—and far more useful if the goal is a quantum network rather than a spectacularly delicate transmission. (China Connected Quantum Memories Across 420 Kilometers of Fiber)

A team led by University of Science and Technology of China researchers Pan Jianwei, Bao Xiaohui and Zhang Qiang reported entanglement between two cold-atom quantum memories across 420 kilometers of optical fiber. According to the accepted Physical Review Letters paper, the team converted photons from the memories into a low-loss telecom wavelength, then stabilized timing and phase across the link. Beyond 230 kilometers, the experiment’s entanglement-generation probability surpassed the theoretical limit for direct transmission without a repeater.

Quantum memories are the waiting rooms of a quantum repeater. They hold information until adjacent links are ready, allowing entanglement to extend across a network. If the architecture scales, intercity quantum links could connect processors, sensors and cryptographic infrastructure without requiring one photon to survive the entire journey.

This remains a two-node experiment, not a functioning repeater chain. The decisive next signal is entanglement swapping—joining separate memory links into a longer connection—with useful rates and fidelity. If that step stalls, 420 kilometers remains an impressive point-to-point result rather than the beginning of a network.

The Wire China’s 2025 “Quantum Panic” treated China’s broader ambitions as a strategic story. This paper provides the more useful evidence: a specific, peer-reviewed networking measurement rather than another declaration of national intent. (China’s “quantum panic” is more about policy than physics)

Noise usually tells quantum engineers that the machine failed to do what it was told. Researchers from the Joint Quantum Institute, the National Institute of Standards and Technology, IBM and collaborating institutions have now shown that, for a narrow machine-learning task, a controlled amount of randomness can help.

The team built a small quantum neural network to classify handwritten digits, then tested it on IBM superconducting hardware and two trapped-ion systems. The researchers report that introducing randomness through quantum measurement improved classification relative to running the network deterministically. Too much randomness hurt performance again; on one deliberately difficult image, the best trapped-ion result was nearly perfect, while IBM’s system reached roughly 90%.

This is not quantum advantage over conventional machine learning. It does show that imperfect quantum hardware can serve as an experimental platform for studying how randomness affects learning—and that noise may occasionally be something to tune rather than merely suppress.

If the effect survives larger datasets, entangling circuits and comparisons with strong classical models, it could create a useful niche for near-term machines that remain nowhere close to full error correction. Failure will be equally clear: the improvement disappears once researchers use better classical baselines or scale beyond carefully chosen examples. The next experiment, not the current accuracy figure, decides which story this becomes.

The Internet Engineering Task Force is turning post-quantum cryptography into protocol machinery, where small details can become global failures. It published revision 09 of its proposal for using standalone ML-KEM in Transport Layer Security 1.3 on July 20. ML-KEM is the National Institute of Standards and Technology’s principal post-quantum method for establishing a shared secret across an untrusted network. (The IETF Revises Standalone ML-KEM for TLS 1.3)

The draft assigns proposed TLS identifiers to ML-KEM-512, ML-KEM-768 and ML-KEM-1024, while adding detailed guidance on key validation, ciphertext-length checks, randomness reuse and side-channel defenses. Those details sound painfully small until one becomes an interoperability failure—or a vulnerability repeated across millions of servers. (The IETF Revises Standalone ML-KEM for TLS 1.3)

The unresolved question is whether ML-KEM should travel alone. All three variants remain marked “Recommended: N,” meaning the IETF has not endorsed standalone deployment as the general default. Many implementers instead pair post-quantum and conventional methods in a hybrid exchange, so an attacker must defeat both. (The IETF Revises Standalone ML-KEM for TLS 1.3)

If standalone ML-KEM advances into an adopted Request for Comments and major TLS libraries converge around it, deployments become simpler and smaller. If consensus fails, hybrid cryptography may become a long-term safety rail rather than a temporary bridge—adding overhead but reducing the risk that one young algorithm can compromise the migration. (The IETF Revises Standalone ML-KEM for TLS 1.3)

That is the operational answer to older general-audience warnings, including Live Science’s 2025 framing that quantum computers could make cryptography obsolete: defenses exist, but turning mathematics into interoperable infrastructure is the real work. It also resolves the unresolved trigger from last week’s issue only partly; revision 09 shows movement, not yet convergence.

DARPA is putting PsiQuantum’s fault-tolerant ambitions under a $125 million microscope. PsiQuantum disclosed an expanded $125 million agreement with the Defense Advanced Research Projects Agency under the Quantum Benchmarking Initiative. The program is designed to test whether commercial architectures can plausibly reach utility-scale fault tolerance—enough reliable, error-corrected computation to solve valuable problems—not merely accumulate impressive physical-qubit totals.

PsiQuantum’s architecture uses photons moving through semiconductor-fabricated circuits. Photons are naturally suited to networking, and the processor itself need not sit near absolute zero. The brutal part is generating, routing and detecting enormous numbers of photons while keeping losses low enough for error correction to work. (DARPA Put Another $125 Million Behind PsiQuantum’s Photonic Bet)

If PsiQuantum’s manufacturing and error-correction assumptions survive DARPA’s evaluation, the company gains something more valuable than another funding announcement: outside validation that a photonic roadmap belongs in the front rank of fault-tolerant contenders. That could redirect government procurement, foundry investment and technical talent toward photonics. (DARPA Put Another $125 Million Behind PsiQuantum’s Photonic Bet)

But the agreement is not a certification. The $125 million figure and program description come from PsiQuantum’s announcement, and the company has not delivered a completed fault-tolerant computer. The observable test is whether DARPA advances the architecture after reviewing loss rates, component yields and error-correction overhead—or whether the roadmap remains expensive PowerPoint with excellent lighting.

CoinDesk’s June account framed the broader U.S. quantum bet at $2 billion and questioned the defense establishment’s ability to evaluate it. The Quantum Benchmarking Initiative is the concrete response: pay specialists to attack vendors’ assumptions before the government starts buying machines built on them. (DARPA Put Another $125 Million Behind PsiQuantum’s Photonic Bet)

Executive orders announce priorities. Budget instructions force agencies to start finding the money.

On July 21, the White House Office of Science and Technology Policy and Office of Management and Budget issued guidance for federal agencies preparing fiscal 2028 research proposals. The memorandum names quantum information science as a strategic priority and directs attention toward quantum materials, photonics, precision measurement, algorithms and post-quantum cryptography.

It also tells agencies to align investments with QC-ADDS, the federal effort to develop a quantum computer capable of scientifically valuable calculations beyond current classical systems. Shared facilities, specialized computing architectures and access spanning universities and industry are part of the requested approach.

If agencies turn that language into measurable programs—with defined workloads, error rates and independent verification—vendors will have to compete on engineering evidence rather than roadmap adjectives. Universities and smaller companies could also gain access to infrastructure that would otherwise remain concentrated inside a few large laboratories.

Failure looks much more familiar: broad priorities enter agency proposals, Congress funds only fragments, and no common benchmark connects the resulting programs. This memorandum is budget guidance, not an appropriation.

CyberScoop’s frequently recirculated warning to federal technology leaders was published in September 2025, not this week. Its migration argument remains relevant, but the new development is that quantum computing and post-quantum cryptography are now written into the federal budget-planning process.

A pair of cold-atom memories whispering across 420 kilometers, a neural network seasoning itself with machine noise, and DARPA standing over a photonic roadmap with a red pen: quantum progress has rarely looked more like three engineering departments refusing to share a building.

Meanwhile, China Southern Power Grid is shopping for post-quantum chips before the IETF has decided whether the cryptography should leave the house without a classical chaperone.

Keep the photons honest.

Forward this to the person who still thinks quantum readiness begins when the quantum computer arrives. (China Southern Power Grid Opens a Tender for Post-Quantum Security Chips)