Magic State Labs founder: AI lets one lab pursue 'hundreds of directions'

Pradeep Niroula, founder and CEO of New York–based Magic State Labs, is building AI systems that can pursue "hundreds of scientific directions at once with minimal guidance" — a shift he compares to moving from playing a violin to conducting an orchestra.

Magic State Labs applies AI to accelerate discovery in quantum information science. Niroula, who studied physics and mathematics at Harvard University and the University of Maryland and previously held research positions at IonQ, Google and JPMorganChase, described his current work as a scaled-up version of his graduate-school routine: reading papers, identifying questions, and testing ideas with mathematics and code. The startup model replaces one researcher's direction with many, orchestrated by machines.

The interview, part of a Q&A series, touches directly on questions R&D managers now face: how AI changes the skill mix inside research teams, what is lost when tools absorb the technical work, and how to train people whose value lies in judgment rather than output.

How does AI change the researcher's job?

Niroula said his attention has shifted from quantum information itself toward meta-science — the study of how scientific discovery happens. The reason is capacity.

"At the startup, I'm trying to build systems that can pursue hundreds of scientific directions at once with minimal guidance," he said. In spirit, the work mirrors graduate school; in scale, it does not.

His orchestra analogy carries a management implication. A conductor, he argued, must still understand the technical details of the music and hold judgment about what it should sound like. He has watched countless hours of Leonard Bernstein conducting, and sees the same texture entering theoretical physics: assembling an ever-growing literature, powerful AI models and human researchers into one discovery process.

With AI models taking on more of the technical load — writing computer simulations, working through mathematical proofs — Niroula argued that scientific taste, or discernment, becomes more central, not less. "We have more capacity to pursue ideas, which makes choosing those ideas all the more consequential," he said.

What does the company actually lose?

The costs he named are concrete. Running a company brings hiring, budget management and fundraising — challenges he said no academic background truly prepares you for, with most learning happening on the job.

He also flagged a subtler loss for researchers whose tools now close the loops: "I also miss the small pleasure of solving a difficult problem." He described spending an afternoon wrestling with an integral or chasing a bug, and the satisfaction that followed. "Those small pleasures have been all but eliminated by AI tools."

On the reward side, he frames his approach as the "gangster" way of doing science: asking audacious questions, exploring improbable directions, and being less intimidated by technical difficulty. AI, in his account, lowers the cost of technical exploration and buys boldness.

What would he tell early-career researchers?

Niroula's advice runs against narrow specialization. He specialized fast — writing papers before graduate school even began — and now says he would have read more history of quantum mechanics, taken more chemistry or biology, and attended more lectures far outside his field.

His diagnosis of the risk is organizational as much as personal: science is a profoundly social project, and even scientists are susceptible to intellectual fashions, particularly when citations, publications and conference panel invitations become the measures of success.

He pointed to Nobel laureate Bill Phillips, who was in the audience at his thesis defence on a highly specialized topic and, in Niroula's recollection, immediately grasped the gist of the problem. That kind of judgment, he said, develops over years of practice.

His summary for the AI era is a portfolio skill, not a technical one. "I believe that the greatest asset you can have as a scientist today is intellectual mobility," he said — the ability to move between disciplines, institutions and levels of abstraction with ease.

Niroula is also a novelist; his fiction debut, The Innkeeper, set in a bed-and-breakfast whose guests arrive from different eras, is due out next fall.