We are currently building the future on antiquated foundations. Let me explain.
When we spin up local AI models or marvel at the latest cloud-based agents, we feel like we are touching the absolute bleeding edge of technology. But in reality, we are just maxing out the physical limits of classical computing.
Welcome to leadprompt.sh // executing leadership from the root. I’m your host, John Collins.
Right now, I run a local AI infrastructure I jokingly name Wintermute on a Framework desktop. It orchestrates local Qwen models, handles vector storage, and automates my workflow. It is an incredibly powerful system. But beneath the software abstractions, it is entirely classical. It is bound by binary logic, silicon wafers, and the physical constraints of traditional transistors. Every model we build today, no matter how many billions of parameters it boasts, is firmly rooted in the Classical AI period.
The Quantum AI period will eventually make our current models look like primitive calculators. But that reality is still decades away, and the delay is not a software problem. It is a fundamental physics challenge.
The Scaling Bottleneck
The central challenge in quantum computing right now is hardware scaling. Specifically, scaling qubits and shielding them from external noise remains the primary hurdle. In classical systems, we relied on Moore's Law to predictably shrink transistors and double processing power year over year. But there is no Moore's Law for quantum computing yet. You cannot simply "add more qubits" to a system safely and expect linear scaling.
Qubits are notoriously fragile. The slightest thermal fluctuation, a stray electromagnetic field, or even cosmic radiation causes decoherence, destroying the quantum state. But the noise isn't just environmental; it is inherently architectural.
Wiring noise is an absolute model killer. In standard superconducting quantum computers, every qubit requires physical wiring for control signals. The more qubits you wire into the quantum computer to scale the machine, the more thermal and electromagnetic noise you introduce into the system. It is a paradoxical engineering trap: the very act of building a more powerful system fundamentally destabilizes it.
Researchers are currently experimenting with microwaves as a viable replacement for these physical wires, attempting to multiplex control signals wirelessly to reduce that physical clutter. But despite these efforts, in my opinion we remain strictly in the "vacuum tube" phase of quantum computing. Think of the massive, room-sized ENIAC machines of the 1940s, that is where our quantum infrastructure sits today.
The Optical Alternative
If the hardware is at the vacuum tube stage, the software layer, for example Quantum AI, is even further behind. You cannot build the software of tomorrow if you do not have the stable hardware of today to compile it on. However, recent international breakthroughs indicate we are finding different paths up the mountain.
In May 2026, a Chinese research team from the University of Science and Technology of China (USTC) brought a groundbreaking machine online named Jiuzhang 4.0. Unlike traditional superconducting machines that require absolute zero temperatures and massive helium dilution refrigerators, Jiuzhang 4.0 is a photonic quantum computer. It uses particles of light, namely photons, that are navigating through a complex optical network.
By utilizing light instead of physical, wired qubits, this architecture entirely bypasses the wiring noise constraints that cripple traditional scaling. Jiuzhang 4.0 utilizes 1,024 squeezed-state inputs and operates across an 8,176-mode interferometric network, capable of completing highly complex Gaussian boson sampling tasks in just 25 microseconds, a calculation that would theoretically take the world's most powerful classical supercomputer over 10^42 years. This proves that stable quantum hardware scaling is actively being solved when we stop trying to force classical wiring paradigms into the quantum realm.
What impact will Quantum have on AI?
So what happens when these stabilized quantum architectures finally collide with artificial intelligence? The impact will rewrite our entire approach to computing. Right now, we train large language models by brute-forcing massive datasets through thousands of sequential GPUs, tweaking weights and biases over months. Quantum AI will discard that sequential brute-force approach entirely. By leveraging superposition and entanglement, quantum neural networks will evaluate vast, multidimensional parameter spaces simultaneously. Training cycles that currently take months and consume megawatts of power could theoretically collapse into minutes. More importantly, Quantum AI will move us past statistical pattern matching. Instead of merely predicting the next token in a sequence, these models will be able to natively simulate reality: from mapping complex molecular structures and discovering new materials, to instantly solving global supply chain routing problems that are mathematically intractable for classical machines.
Execute with What You Have
As leaders and engineers, we have to execute in the reality we currently inhabit. That means building, optimizing, and deploying classical AI models. Keep refining your local infrastructure, keep tuning your RAG pipelines, and keep shipping code.
But maintain a clear-eyed view of the horizon. The AI we use today is the peak of the classical era. The moment quantum hardware stabilizes enough to support true Quantum AI, the entire technology stack we rely on today will be rewritten from the root.
Until next time, keep executing from the root.
References:
- Liu, Z. (May 15, 2026). "Does China's Jiuzhang 4.0 computer herald the age of quantum supremacy?". South China Morning Post.
- CGTN. (May 14, 2026). "China's new quantum computing prototype 'Jiuzhang 4.0' sets world record".
- Pan, J.-W., & Lu, C.-Y. (2026). Ongoing quantum research on the Jiuzhang 4.0 1,024-input, 8,176-mode interferometric network at the University of Science and Technology of China (USTC).
Download audio
File details: 9.7 MB MP3, 6 mins 45 secs duration.
Title music is "Apparent Solution" by Brendon Moeller, licensed via www.epidemicsound.com
