The Singularity Monthly: Quantum Crunch
In 1994, computer scientist Peter Shor devised a program no machine on Earth could run. Shor’s algorithm, as it’s now known, was written for a quantum computer—a machine that exploits the weird laws of quantum mechanics to process information. The algorithm could perform a particular type of calculation that would take classical computers billions of years to complete. But this wasn’t just proof that quantum computing could outrun classical machines on some abstract problem. The same calculations also happen to secure the entire digital world: A quantum computer running Shor’s algorithm would shatter modern cryptography.
What might that mean in practice? Nothing good. Every email, text, and click would be wide open. Social security numbers and credit card information would pour from corporate servers. Bitcoin would be no better. Anything digital, from health to national security, could be hacked—and we’ve spent decades ensuring everything is digital.
But originally Shor’s algorithm was a score with no orchestra to play it. Even when researchers finally built the first quantum computer in 1998, it had only two processing components, or qubits—far short of the millions experts estimated we'd need to run the algorithm. Since then, breakthroughs in scaling, error correction, and algorithmic design have whittled away at the years left until Shor’s quantum apocalypse (also known as Q-day). Quantum computers have grown in size to over a thousand qubits, while at the same time, researchers have been devising clever error-correction schemes and tinkering with Shor’s algorithm to make it more efficient.
In 2019, Google's Craig Gidney coauthored a paper showing the number of qubits needed to break RSA, a widely used form of encryption, could be reduced from 170 million to 20 million. Then in 2025, Gidney again slashed the number, this time to just under a million. That’s over two orders of magnitude lopped off best estimates in just six years. Now, in the span of a few months, three new papers have brought an internet-breaking quantum computer to within spitting distance of today’s machines.
We highlighted the first paper in last month’s newsletter. A team of researchers in Australia led by Paul Webster at Iceberg Quantum wrote that, if you assume the possibility of greater interconnectivity between qubits, it would take a 98,000-qubit machine about a month to break RSA. One with 471,000 qubits could do it in a day. Building such a machine—which, like in Gidney’s research, assumes the superconducting qubits favored by Google and IBM—would be no small feat. But IBM has said it’s working on the hardware needed to realize this kind of connectivity.
Several weeks later, Gidney’s team published a new white paper. This time they focused on elliptic curve cryptography (ECC), another popular form of encryption notably also used in cryptocurrency. The researchers demonstrated a way to break ECC that was 10 times more efficient than prior efforts. Most cryptocurrencies, including Bitcoin, would fall to a machine with 500,000 qubits in minutes.
The final paper, published by Caltech researchers, focused on quantum computers that use neutral atoms for qubits. These computers are some of the biggest yet. Atom Computing has a working machine with over 1,000 qubits, and Caltech built an array—that is, a lattice of atoms that has yet to run algorithms—with just over 6,000 qubits. Neutral-atom computers offer more interconnectivity because the atoms can be rearranged on the chip, but they’re also slower than their superconducting cousins.
In the paper, the team used AI to develop a more efficient error-correction strategy. They then pitted a neutral-atom computer against RSA and ECC in simulations. With just 10,000 qubits, they found the computer could break RSA in a century (an enormous step down from billions of years, but nowhere near practical) and ECC in about three years (still impractical but enough to make Bitcoin users sweat). When they upped the sizes to 100,000 and 26,000 qubits, respectively, the times it would take to crack encryption fell to three months (RSA) and a few days (ECC).
Running Shor’s algorithm was deemed well beyond reach not long ago, but it now looks plausible for machines on the horizon. Both IBM and Google are targeting quantum computers at these scales within the decade. More importantly, the trends in error correction and algorithmic improvement are likely to continue. There’s no Y2K-like date for cryptography’s demise, but the pace of its approach is accelerating.
Now, there’s good news and bad news. Because the industry has known what Shor’s algorithm portends for so long, there’s a solution: Post-quantum cryptography. In 2024, the US National Institute of Standards and Technology (NIST) approved three algorithms to repel quantum attacks. Early adopters include Signal and Apple. But here’s the bad news: Winning wider adoption may be a bit like trying to steer a glacier. NIST’s official deadline is 2035. But just before Gidney’s paper dropped, Google bumped up its own adoption target to 2029. The message: The time to get serious is now. If post-quantum cryptography catches on, Q-day could be a yawner.
MORE NEWS | From the Future
Robotics startup sees “early signs” of generalization in new AI model.
Startup Physical Intelligence says the AI controlling its robots can now perform tasks it wasn’t trained on. In a blog post, the team describes a test where, with careful prompting, the robot learned to operate an air fryer, which the researchers struggled to find in its data set. The model, they say, transfers skills learned for unrelated tasks to handle new ones and is a small step toward LLM-like generalization. “The last few months have been the first time where I’m genuinely surprised,” Physical Intelligence’s Ashwin Balakrishna told TechCrunch. “I just bought a gear set randomly and asked the robot, ‘Hey, can you rotate this gear?’ And it just worked.”
Scientists “boot up” zombie cells by giving dead bacteria new genomes.
In a bioXriv preprint, a team at the J. Craig Venter Institute dosed bacteria with a lethal chemotherapy drug that destroyed their DNA. The scientists then transplanted synthetic genomes from another species of bacteria into the functionally dead cells. Some of the resurrected “zombie” cells grew and divided, and their offspring inherited the synthetic genomes. This makes them the first synthetic cells assembled from “nonliving” parts, according to the team. The work “blurs the line between life and death” and could advance efforts in synthetic biology to engineer cells that do valuable work, like producing fuels, drugs, or new materials.
Internet speed record could stream 50,000,000 movies at the same time.
Researchers seem to break the internet speed record at least once a year. But it's usually in the lab with novel equipment, and we’d have to rip up and replace existing cables to make it happen in the real world. Recently, however, researchers showed they could boost data speeds by nearly 10 times on existing cables in London. The team recorded speeds of 450 terabits per second by drastically expanding the frequencies of light used to encode data. The new bandwidth could meet growing demand, especially from AI, and be up and running in five years, they said.
CALENDAR
Upcoming Events
OCTOBER 25-29 Singularity Executive Program | Silicon Valley, California
MAY 21 The Discussion Series | Future Proof: How to Lead in the Age of AI with Dr. Michael Housman | online webinar and Q&A
Thanks for reading. We hope you enjoyed this month's updates and found something to inspire you on your exponential journey.
See you next month!
The Singularity Team
.jpg)