AI agents paired with today's quantum computers are dramatically accelerating research timelines, compressing months of work into days or weeks. Mykola Maksymenko, a quantum physicist and co-founder of Haiqu, a quantum software startup, recently told his team he believes artificial general intelligence (AGI) has arrived, not because quantum computers have suddenly become perfect, but because intelligent AI systems can now navigate their complexity and unlock practical value from existing hardware.

What's Actually Happening Inside These AI-Quantum Systems?

Haiqu launched what it calls an agentic quantum operating system in May. The concept is straightforward: AI research agents break down hard problems into smaller pieces, identify which quantum algorithms might apply, and then use Haiqu's software to optimize and run the job on actual quantum hardware. The result is a dramatic compression of research timelines.

Maksymenko's own work provides a concrete example. A piece of his PhD research that took approximately six months to complete was recently rebuilt by an AI agent with access to modern large language models (LLMs), which are AI systems trained on vast amounts of text data, and a quantum computer. The agent finished the work over a single weekend and even found a bug in one of his formulas.

"A couple of months back, I came to a meeting with our team and said, 'Guys, I think AGI is here,'" Maksymenko told me in a recent NEXT with John Koetsier podcast.

Mykola Maksymenko, Co-founder and CTO at Haiqu

Maksymenko quickly clarified his claim. He describes this as "weak AGI in some sense" that is "probably not completely self-consistent, but you can already use it for many good tools." The key insight is that stacking frontier AI models, agentic systems (AI that can break problems into steps and solve them autonomously), and current-generation quantum computers together might shortcut the decade-long wait for quantum computing to become genuinely useful.

How Are These Speedups Translating Into Real Work?

The acceleration Maksymenko describes is not marginal. In his personal case, he estimates the system allows him to speed up work "probably from half a year to a week." One research partner using the system has been getting "promising, very complex results" after "a couple of weeks instead of working for a year".

Haiqu's launch materials make an even more striking claim: a molecular dynamics simulation that previously cost "$30,000 and more than nine hours" dropped to "about $25 in roughly 30 seconds." While independent verification is needed, this kind of cost curve fundamentally changes who can do advanced, cutting-edge work and how quickly research can progress.

As a side project, Maksymenko pointed the agentic system at genomics, a field outside his expertise. Earlier this year, a team funded through Wellcome Leap's Quantum for Bio program, including researchers from the Wellcome Sanger Institute and the universities of Oxford, Cambridge, and Melbourne, announced they had loaded a complete genome from the hepatitis D virus onto an IBM quantum computer for the first time. Maksymenko downloaded the publicly available genome data and gave it to Haiqu's agentic system. Working largely from the announcement itself, the agents guessed what the researchers had probably done, which algorithms could be used, and how to encode it with Haiqu's tools.

"I was surprisingly able to reproduce the results pretty quickly," he says.

Mykola Maksymenko, Co-founder and CTO at Haiqu

The system loaded the genome and then made small mutations and tracked them as they happened. A physicist with no genomics background replicated a frontier biological research project almost as an afterthought. If Haiqu can replicate this pattern across different projects and fields of study, it represents a massive tool for scientific acceleration.

Why Quantum Utility Matters More Than Quantum Supremacy?

For years, the quantum industry's primary milestone has been "quantum supremacy," the point at which a quantum computer performs tasks no classical computer realistically can in a reasonable timeframe. Maksymenko is blunt about what those demonstrations actually represent.

"We know that those experiments were very much artificial, in some sense, that they were very fine-tuned to a specific mathematical problem which has nothing to do with useful reality," he stated.

Mykola Maksymenko, Co-founder and CTO at Haiqu

The real question, Maksymenko argues, is whether you attack problems that actually matter. That's "quantum utility," and it's where Haiqu is aiming right now. The company is focusing on optimization, quantum machine learning, and hybrid high-performance computing tasks like "diagonalizing huge matrices," which means running massive matrix calculations used to predict how molecules and materials behave.

Maksymenko says quantum utility can happen today on the limited hardware we have, with "a hundred qubits to two hundred qubits, not necessarily logical." In other words, noisy physical qubits, not the error-corrected logical qubits most industry roadmaps say we need before quantum gets serious.

Maksymenko

Steps to Unlock Quantum Utility Today

Rather than waiting for perfect quantum hardware, researchers and companies can begin extracting value from current systems by focusing on specific problem domains and leveraging AI to navigate complexity:

• Target Natural Quantum Problems: Focus on quantum dynamics and quantum chemistry, which are naturally suited to quantum machines because there is no heavy lifting required to map a quantum problem onto classical hardware, and calculations take less time and energy.
• Use AI Agents as Intermediaries: Deploy agentic AI systems to break down complex problems, identify applicable quantum algorithms, and optimize code for current hardware, dramatically reducing the expertise barrier for researchers.
• Combine Hybrid Approaches: Stack frontier AI models with current-generation quantum computers to tackle optimization, machine learning, and high-performance computing tasks that benefit from quantum acceleration without requiring fault-tolerant quantum computers.

One example Haiqu demonstrated: the company used AI to create an algorithm simulating neutron scattering on magnetic compounds. The result, Maksymenko says, "was almost on par with what I did in academia, like 10 years ago, on the largest supercomputers".

Maksymenko

"I would say quantum dynamics and quantum chemistry are some of the low-hanging fruits," Maksymenko added. The goal is "to demonstrate that we can run interesting, large-scale applications of the frontier scale already today. So we don't need to wait for 10 years from now for fault-tolerant quantum computers".

What Does This Mean for the Quantum Computing Timeline?

Maksymenko compares the current state of quantum computing to classical computing's early days. "In the past, I compared it to the age of classical computers in the 1940s and 1950s," he explained. "You needed a lot of expertise to manipulate the valves and computers with switches." That was quantum computing as recently as six months ago, in his telling: expensive and extremely hard to run anything on.

Now an agentic layer can handle much of the low-level complexity. In some cases, the agents have found ways to use Haiqu's own software that the company didn't intend, speeding up workflows with different combinations of smaller subroutines. This also lowers the bar for who gets to do science. One of Haiqu's users now sets an agent loose on semi-random back-burner ideas over the weekend, the kind of ideas you'd never normally spend grant money or graduate student years on.

"Now you can just do it almost for free. And sometimes these small ideas can actually lead you to interesting results, which can then probably become even a major research avenue," Maksymenko noted.

Mykola Maksymenko, Co-founder and CTO at Haiqu

Haiqu raised an $11 million seed round in January led by Primary Venture Partners, signaling investor confidence in this approach. The company's thesis is that quantum utility doesn't require waiting for perfect hardware; it requires the right software layer and AI guidance to extract value from machines we already have.

Of course, quantum computing has a long history of overpromising. But Maksymenko's claim is narrower and more testable: not that quantum is ready for everything, but that there are specific pockets where today's machines, driven well and used in conjunction with modern LLMs and agentic AI, beat the alternative. If that holds up under scrutiny, it could reshape the quantum computing timeline from a decade-long wait to immediate, incremental progress.