Chief Quantum Officer

The Executive’s Evidence Diet

In 2019, Craig Gidney and Martin Ekerå published an estimate that factoring a 2048-bit RSA key would take a quantum computer with twenty million noisy qubits, running for eight hours. In May 2025, Gidney published a revision putting it at fewer than a million noisy qubits, running for under a week.

The physical assumptions did not change between the two papers. Same grid, same gate error rate, same surface code cycle time. A twentyfold reduction in the hardware required to break the cryptography most of the world depends on, and mathematics rather than machines accounts for all of it.

That is the most decision-relevant quantum development of the past six years, and an executive reading nothing but hardware announcements missed it completely. Gidney posted it as a preprint on arXiv, with no launch event and no vendor promoting it.

Reading more does not help

The volume will only grow: hundreds of companies, thousands of papers a year, and a press release cycle in which vendors label every incremental result a milestone.

The instinct is to read more, which fails for a structural reason. Most quantum publications report progress toward a capability, not a change in what is possible, so an executive has nothing to decide on. When a qubit count goes from 1,000 to 1,200, an executive does nothing differently on Monday. When the estimated cost of breaking RSA falls by a factor of twenty, the executive has to revisit the migration deadline, the risk register, and possibly the budget.

The useful discipline isn’t reading widely. It’s knowing the few document types that can change a decision and watching them closely enough to notice a new one.

Four things that can change a decision

Resource estimates

The category the Gidney paper belongs to, and the highest-yield reading in the entire field for anyone with a security mandate.

In these papers, researchers work out what it would actually cost, in physical qubits and time, to break a specific cryptosystem under stated assumptions. There are only a handful of serious ones per year. Each is checkable, each states its assumptions explicitly, and each is directly convertible into a planning input.

Two habits make them usable without a physics background. Read the assumptions section first, since the headline number is meaningless without it and the authors write it in plain language. And watch for the author’s own statement of limits. Gidney wrote that he saw no way to reduce the qubit count by another order of magnitude without changing the physical assumptions, and an executive learns more from that sort of admission than from the abstract.

Standards and regulatory documents

Dated, binding, tedious, and the only category with a deadline attached.

NIST publications, national transition guidance, and sector regulator advisories determine what an organization must do and by when. They are also the most under-read source in the field, precisely because they are dull. An executive who has read the actual guidance rather than a summary of it will be better informed than nearly everyone in the room.

Primary results with stated assumptions

The paper behind an announcement, not the press release.

Check whether the claim states the conditions under which it is true. If the authors give error rates, the size of the instance, and how a classical computer performs on the same problem, you have a result. If they specify none of those, you have marketing with a citation attached.

Where it is published is less important than most people assume: the Gidney revision was a preprint and still the most consequential document of its year.

Money and procurement

What large buyers actually purchase is a better signal than what anyone says.

From government procurement programmes, national funding commitments, and where serious capital goes, you can see which parts of the field investors now consider fundable rather than merely interesting. This category requires no specialist reading at all, and technical people ignore it for exactly that reason.

What to leave unread

Qubit count announcements, unless accompanied by error rates and a workload.

Roadmap updates from vendors, other than to note what was delivered against the previous roadmap, the only genuinely informative part of one.

Anything with the words breakthrough or milestone in the headline and no numbers in the body.

Market forecasts predicting the size of the quantum industry a decade out. Nobody has ever been right about one. Analysts produce them to be quoted rather than to be accurate, and an executive can ignore them for every decision this year.

People as instruments

A small network of people who read different things is worth more than any personal reading programme, and the trick is choosing them for coverage rather than seniority.

Three or four people is enough. Somebody academic who reads the algorithms literature, somebody in a security function at a peer organization facing the same regulatory pressure, and somebody on the policy side. The relationship works when it is reciprocal and specific, so send them the two things a year they would want and ask for the same rather than requesting a general briefing.

One caution: If all of them are connected to one vendor, you have one source, not four.

The time budget

Two hours a week, honestly protected, is enough to run all of this.

That is roughly one hour on the four categories above, mostly skimming and occasionally reading properly, and one hour on the network. Anyone claiming an executive needs more is describing a research job, and that isn’t the job.

Executives fail at this through irregularity, not volume. Two hours every week is better than eight hours once a month, since the value comes from noticing a new result early rather than from depth of coverage. Most of the failures I have seen came from executives reading in bursts after something had already gone wrong, not from reading too little.

Whether it is working

Use one test. When the next significant result appears, does the organization hear about it from its own quantum leadership, or from a vendor, a journalist, or a board member forwarding a link?

The Gidney revision is the fair example here. After its publication in May 2025, every organization with data that must stay secret for years had to revise its planning assumptions, and a great many executives learned about it months later at a conference. The information was free, public, and available on the day.

There is an old saying in cryptography that attacks always get better. It has been true of quantum factoring across the past decade, which is why this reading has to be continuous rather than periodic. We are not waiting for a machine to be built. We are waiting to find out how much machine it will take, and researchers have already lowered that number once.