Issue #20 | September 1, 2026

(The headliner’s in Tangent this week… all the way at the bottom)

S — Status Report (The Intelligence)

• Crypto4A got FIPS 140-3 Level 3 validation for its QASM module on August 20, certificate 5497, which they call a world first for an HSM supporting all the NIST-approved post-quantum algorithms. The quantum-safe version of the box your keys actually live in finally has paperwork an auditor will accept.

• GSA and Treasury both moved on August 24. Treasury’s Quantum-Readiness Task Force, announced citing EO 14412, made third-party and vendor readiness one of its three workstreams. GSA went at the doors, expanding its FIPS 201 lab to evaluate quantum-resistant solutions for federal buildings, alongside the identity work it runs under OMB M-26-15. The questionnaire and the badge reader, same week.

• Quantum eMotion submitted its eCore-Q module for NIST SP 800-90B entropy validation on August 27. Submission isn’t validation and the review can go either way. Watch it anyway: everything else stands on the generator, and Debian shipped a broken one for 20 months, with every SSH key generated on an affected machine coming out of a pool of 32,767.

H — How It Works: finding crypto that isn’t wearing a name tag

Your cryptographic inventory is a list of names.

That’s not a dig, it’s just what the tooling reads: symbol names inside the binary, package names in the manifest, and the filenames of whatever libraries you linked against. Some of them go into the compiled binary too, and they’re looking for the same thing in there, “what is this called,” then guessing the crypto from the answer.

So take that same code and compile it with optimization on, strip the symbols, static-link it, and the names are gone while the code runs exactly the same. Your inventory goes very, very quiet about something that’s still very much running…

Which raises the obvious question: how do you find a thing that won’t tell you its name???

Here’s the useful bit. ML-KEM does its arithmetic against a fixed table of numbers, a number-theoretic transform table, and those numbers come straight out of FIPS 203, the published standard. The programmer doesn’t choose ‘em and the compiler can’t touch ‘em, because altering a constant changes the answer the program gives. The paper’s own line for it is the best one: the function name may be removed, whereas the constant 3329 cannot.

What the heck does that mean in normal language?

Picture a warehouse. You count what’s in it by walking the aisles and reading the labels on the boxes. Simple, fast, works greatttt, right up until you find out the labels have been peeling off in shipping for three years and nobody mentioned it. Your clipboard says the place is half empty, but the warehouse is FULL, and it’s been full the whole time.

So you stop reading labels and put the boxes on a scale instead, because a crate of the same thing always weighs the same, and this one had its weight printed in a public standard before anybody shipped it.

The label comes off. The weight doesn’t.

A paper published August 25 does precisely this, and I’ll flag it up front: single author, no affiliation listed, so treat it as a finding to go check rather than gospel. It scans read-only data for the tables instead of hunting names. Pointed at 6,224 binaries on one production Linux box, it found 24 paths resolving to 12 distinct programs carrying ML-KEM, none of them on anybody’s inventory. Including the OpenSSH key-exchange program and the container stack.

And here’s the part that got me: they got there two different ways, and neither one is sneaking. OpenSSH decided. Release 10.0, April 9 2025, a public note saying mlkem768x25519-sha256 “is now used by default for key agreement.” 16 months, sitting in a changelog anybody could read.

The other group is containerd, docker, runc and ctr, and they decided nothing at all. They’re carrying the Go standard library’s ML-KEM, which the paper says arrived “through the toolchain rather than through a deliberate decision by the authors of each project.” One group announced it, the other never chose it, and an inventory built from names missed both, because names are the one thing a stripped binary stops having.

So the estate you’re pretty sure hasn’t started migrating… may have already started migrating, in 12 places, every one of them documented somewhere nobody looked.

Go ask whoever owns your inventory what it does when the symbols are stripped. Spoiler: it says everything’s gucci.

I — Investment Watch

• Pasqal closed its combination with Bleichroeder Acquisition Corp. II on August 27 and opened on Nasdaq as PSQL the next day, finishing up 95.2% at $19.11 on a roughly $2 billion valuation, with about $360 million of cash at closing. Another neutral-atom qubit shop taking the SPAC door to the public markets.

• Canada committed CAD $195 million (about $140.2M USD) to Xanadu on August 28 through the Strategic Response Fund, anchoring a CAD $893 million project to build “Inception,” a 158,000 square foot photonics manufacturing facility in Toronto. It’ll create 275 jobs, and they’re calling it the largest of its kind in Canadian history.

• NSF put more than $290 million into eight Quantum Leap Challenge Institutes on August 25, roughly $28M to $37M each over five years across 36 universities in 19 states, aimed at fault tolerance, error correction, networking, sensing and simulation, which is the list NSF published. Between this and Ottawa, governments put up more cash this fortnight than the quarter’s biggest listing raised.

F — Field Notes

In every 2026 cryptographic discovery conversation, the first question is which scanner to run. Reasonable question! It also skips a step, and the skipped step is the one I can’t stop chewing on.

Every scanner tells you what it found. Not one of them tells you what it missed.

And the accuracy numbers don’t save you, because no two of them were measured against the same thing. Each vendor built its own ground truth, on its own corpus, graded by its own people. So “98% recall” from one tool and “94% recall” from another aren’t two points on a line… they’re two separate experiments that happen to share a vocabulary word.

Which means when two tools disagree about the same file, there’s no referee. You go with the one whose report you like, or the one you already paid for.

So starting this month I’m fixing that the slow, stupid way. Six months of hand-labelling cryptography in real code bases across five ecosystems, Python first, then Go, Java, C#, and C/C++, with no scanner anywhere near it. Every call written down with the evidence behind it, including the ones where I was wrong and had to walk it back.

Day one produced ZERO labels, because I’d picked a file that teaches a lesson I hadn’t gotten to yet, which is ~its own kind of finding~.

And the reversals are the valuable part, which genuinely surprised me. Every time I label something one way on Monday and change my mind on Thursday, that argument is a tiny version of the exact fight two scanners have with each other, except this time I can see both sides, because both sides were me.

I’m doing it so that when a client’s tool and a specialist’s tool disagree about one file, I can open the file and say which one was right, instead of forwarding somebody else’s confidence with my name on the cover. If you’re going to sit in the middle and call yourself vendor-neutral, you should probably be able to check the vendors.

T — The Tangent: We Already Ran This Play, and It Was Called an SBOM

May 12, 2021. In the wake of SolarWinds, the White House signs Executive Order 14028. Buried in a long doc about federal cybersecurity lies a requirement that sounds boring and absolutely is not: Commerce gets 60 days to define the minimum contents of a Software Bill of Materials, and agencies are told to start requiring one from the people who sell them software. A list of every component inside the thing you’re buying.

The logic was clean. You can’t defend what you can’t enumerate. Give everybody a parts list, and the next time some component blows up, you’ll know by lunchtime whether you’re holding one.

December 2021. Seven months later, and it’s Log4Shell.

A logging library that approximately every Java application on earth pulls in, sitting several dependencies deep in stacks nobody had ever drawn a picture of. And the question every security team on the planet asked that weekend was the exact question the executive order had been written to answer: am I running this?!?!

Mostly, they couldn’t say. People spent the holidays grepping.

Worth knowing how long the clock had been running before any of that, too. Chen Zhaojun of Alibaba’s cloud security team reported the bug to Apache on November 24, weeks earlier. So by the time the question went public, most organizations could say only that they’d been exposed for an unknown period, to an unknown degree, in an unknown number of places, and an inventory was supposed to retire at least two of those three.

Then, in January 2026, OMB rescinded it. Memo M-26-05 pulled both M-22-18 and M-23-16 and called the old regime “unproven and burdensome software accounting processes that prioritized compliance over genuine security investments.” Agencies can still ask for an SBOM when they judge the risk warrants one, but the government-wide requirement is gone. That didn’t settle the argument, and both sides still have a case.

For: the teams that had actually built inventories reported answering in days while everyone else spent the holidays grepping, and the gap was visible from outside. Asking a vendor for a parts list went from a weird request to a normal contract line. That’s a real shift and it happened fastttt.

Against: an SBOM lists what you declared, not what’s running. It’s generated at build time, from the manifest, by the same build that made the artifact. Which makes it excellent at describing the parts somebody wrote down, and blind to everything arriving underneath them. Plenty of teams generate one every single build, file it, and have never once opened it.

Now go back and look at those 12 programs.

Not one of them shows up on an SBOM. Nobody at those organizations declared ML-KEM, because nobody there chose it. OpenSSH chose it, and the Go toolchain chose it for containerd, and either way it landed three layers below anything a manifest records. The parts list is accurate and the parts list is wrong, simultaneously, which is the most annoying possible outcome.

And the crypto version of this is being built right now. It’s called a CBOM. Same idea, same promise, pointed at algorithms instead of packages. It’ll be genuinely useful, and any CBOM generated the way most SBOMs are, from the manifest at build time, inherits the same hole on day one. Worth sitting with: the same office that scrapped the software parts list in January issued the post-quantum migration memo in June, and the executive order behind it has CISA defining CBOM minimum elements by next spring.

So it’s worth deciding now, while yours is still a blank document, whether you want a list of what your developers said they used or a list of what’s actually on the disk.

Those are not the same list. They weren’t in 2021 either.

See ya in two weeks!

Addie LaMarr

Founder, LaMarr Labs

⚛️

Issue #19 | August 18, 2026

S — Status Report (The Intelligence)

• On August 6 somebody claimed a fast quantum method for cracking the math post-quantum encryption’s built on. By August 15 it was dead. The preprint went after an obscure problem called the Dihedral Coset Problem, which matters only because a 2004 result from Oded Regev, sharpened in 2018, ties it to lattice problems and Learning With Errors, the trick of hiding a secret inside deliberately noisy equations, which ML-KEM and ML-DSA both rest on. The author called it a preliminary draft. 9 days later three researchers published a proof that the method can’t work, with code attached so a machine could verify it rather than anyone taking their word. No real-world parameter set was ever attacked. The full 9 days, and the strangest detail in them, are in the Tangent.

• QuSecure’s post-quantum platform went onto Carahsoft’s GSA Schedule on August 11, which reads like paperwork and is actually the mechanism. Federal systems have to hit December 31, 2030 for key establishment and 2031 for signatures under NSM-10 and the executive orders on top of it. Anything an agency can’t put on a purchase order by 2029 won’t be running by 2030. Procurement’s a deadline too.

• Pasqal trapped four rubidium atoms on August 10 using laser light generated by a photonic chip instead of a bench full of bulk optics, holding them about 27.5 seconds, which matches what its existing systems do. The company puts the eventual footprint reduction at up to 50x and is aiming at 10,000 atoms. Zero movement on the math this fortnight, and a real step on the hardware.

H — How It Works: why a paper that never mentions your encryption can still be about your encryption

You already know this shape from your day scam job. You ship an app. Your app imports a library. That library imports another library. That one pulls in a tiny package maintained by one guy who hasn’t logged in since 2019. Nobody ever attacks your app. Somebody gets to the guy, and by Tuesday your app is the incident.

Cryptography has a dependency tree too. The difference is that cryptographers built theirs on purpose.

Translation pls:

When somebody designs an encryption scheme, they don’t just assert it’s strong and hope for the best. Instead, they prove a chain: if you can break my scheme, then you can solve this specific math problem. Then they point at the math problem and say, everyone go hit THAT for 20 years. Concentrating the world’s cryptanalysts on a handful of hard problems, instead of scattering them across a thousand individual schemes, is the entire strategy. It’s why the NIST competition looked like a demolition derby on purpose.

So ML-KEM’s security reads like an import statement. ML-KEM depends on a problem called Module-LWE. Module-LWE depends on certain lattice problems staying difficult. And back in 2004, Oded Regev proved another entry into that tree: solve something called the Dihedral Coset Problem on a quantum computer, and you can solve those lattice problems. Brakerski, Kirshanova, Stehlé and Wen tightened the connection in 2018.

Nobody outside a math department has ever heard of the Dihedral Coset Problem. It’s sitting way down in the tree, several levels below anything with a FIPS number on it. And that’s precisely where this month’s paper landed.

Which is why the real description of the last 2 weeks is that nobody attacked ML-KEM at all. The paper never writes down an ML-KEM parameter or costs an attack against a deployed system, and the author says so himself. Somebody took a swing at a dependency four levels down, and if it had connected, everything sitting above it would have come down together.

Spoiler alert: It didn’t connect. And people paid attention anyway, because they can read a dependency tree. Many over reacted with claims that it was the end for PQC itself… per usual.

So next time a headline says a paper “threatens” your encryption, the useful question is how far down the tree did this land, and did anyone walk it back up to a real parameter set??? Almost nothing ever does. This one stayed four levels below the nearest FIPS number.

I — Investment Watch

• IonQ posted record Q2 revenue of $80.1 million on August 5, up 287% year over year, and raised full-year guidance to $280–290 million while sitting on $3.0 billion in cash. That guidance excludes any contribution from SkyWater, the chip fab it just finished buying. Vertical integration is the bet.

• New York opened a $60 million RFP on August 13 for up to four Regional Quantum Technology Commercialization Hubs, up to $15 million each, applications due October 14. State money chasing the gap between a university lab and a product.

• QpiAI opened an eight-inch quantum chip foundry in Bengaluru on August 17, roughly $20–25 million spent so far with another $10–15 million planned, currently fabricating flip-chip superconducting processors up to 128 qubits and targeting 10,000 by 2027. The capital’s moving into manufacturing!

F — Field Notes

Nobody on your team is going to refuse the migration. That’s the part that makes it so hard to see coming, and Homer wrote the mechanism down roughly 2,700 years ago.

Quick aside first: I went and saw the Odyssey this weekend. I’d read it in high school like everybody else, and watching it is a completely different animal. If you haven’t gone, go!!!

With Odysseus gone 20 years, a hundred suitors occupied Penelope’s house and pressed her to choose one of them. She told them she’d choose as soon as she’d finished weaving a burial shroud for her father-in-law Laertes, a duty none of them could argue against. She wove at the loom every day, and every night by torchlight she unpicked the day’s work. 3 years passed this way. The suitors had no complaint available to them, because the work was visibly underway the entire time. It ended when one of her maids told them what she was doing. (Odyssey II.93-110 and XIX.137-156.)

Engineers almost never refuse a cryptographic migration. Refusing is costly and it requires saying something out loud that’s difficult to say, usually that the change threatens hard-won expertise or reads as a verdict on a system they built. What comes back instead is a request for more analysis, a dependency that has to be resolved first, and a well-argued case for waiting until the standards settle. Each of those is defensible on its own. 3 years of them is a shroud.

And Penelope wasn’t being difficult or lying about the weaving. She’d been handed a question she couldn’t answer safely, and this was the only response available to her. The engineer who keeps coming back with ‘one more dependency’ is doing the same thing, and solving a different problem than the one you handed them.

The moral of the story’s this: someone who can’t refuse openly will produce work that’s always in progress and never finished.

T — The Tangent: the third assassination attempt

Three times in 28 months somebody has claimed to kill lattice cryptography. I’ve now written about all three, and the pattern’s gotten weird.

Attempts one and two, briefly. In April 2024 Yilei Chen at Tsinghua published a quantum algorithm that, had it worked, would have taken the entire lattice family with it. Two cryptographers located the flaw within 8 days and Chen conceded it in public. In May 2026 Chevignard, Shen and Schrottenloher claimed something far narrower, a speedup on one component of one attack, requiring quantum memory that doesn’t exist, and their own paper cautioned it might never be realizable. I wrote that one up in May. Neither came anywhere near a live parameter set.

August 2026. And here’s where it gets good. Daniel R. Simon, now in Amazon’s cryptography group, posts a preliminary draft claiming a quantum algorithm that cracks the Dihedral Coset Problem fast enough to matter. Same Daniel Simon whose 1994 algorithm was the first hard proof that a quantum computer could run exponentially faster than any classical computer, on a problem built for the purpose. Peter Shor read it, reworked the same idea to run on ordinary whole numbers, and out came the algorithm that put every one of us in this business. The man is, in a real sense, the reason your inbox has the word “post-quantum” in it.

9 days later Aparna Gupte at MIT, Seyoon Ragavan at Google Quantum AI and MIT, and Mark Zhandry at Google Quantum AI and Stanford publish a note titled, with no ceremony whatsoever, “The ePrint:2026/1591 Quantum Algorithm Does Not Solve DCP.” Their emphasis: “we are showing directly that the algorithm cannot possibly work.” Then they released Lean 4 code so you don’t have to take their word for it. Simon has revised the draft since and says he’s evaluating the result.

Now flip to Simon’s acknowledgments page, because this is the part I can’t stop thinking about. He thanks five people for useful discussions. One of them is Seyoon Ragavan, who then co-authored the refutation. Another is Thomas Vidick, who helped kill attempt number one in 2024.

Two ways to read that. The generous read: this is a field working exactly as designed, where the people who help you sharpen an idea are the same people who’ll publish why it fails, in public, in under 2 weeks, with machine-checkable proofs, and nobody takes it personally. Try getting that turnaround out of any other industry.

The other read is that the same dozen names keep appearing on both sides of every attempt, which means the security of a meaningful fraction of the world’s future encryption currently rests on a group of people who’d fit in one conference room. The field isn’t casual about that. This work is genuinely, brutally hard, and the number of people on earth who can do it at this level is small enough to be a planning assumption.

Three swings, three misses, defense up 3-0. And the reason it’s 3-0 is roughly a dozen people reading fast.

See ya in two weeks!

Addie LaMarr

Founder, LaMarr Labs

⚛️

(Headliner’s in Field Notes for the third issue running. I’ll stop.)

S — Status Report (The Intelligence)

• On July 28 somebody posted an improved attack on a post-quantum signature algorithm called HAWK to NIST’s public mailing list. Another cryptographer replied about 5 hours later: “It checks out independently for me.” The next day HAWK’s own authors pulled it out of standardization, and NIST updated its page to say so. The discovery came from an AI model, Anthropic’s Claude Mythos Preview, which ran about 60 hours at roughly $100,000 and found a hidden symmetry in HAWK’s math that halves its effective key size. Before you panic: HAWK was a candidate in a side competition, never standardized and never deployed. Your stuff is fine. The interesting part’s in the Tangent.

• DigiCert’s second annual survey landed July 23 and it’s grim in a very specific way. Of 1,001 IT and security decision-makers across the US, UK and Australia, 87% say they’re planning, testing or implementing post-quantum crypto, while 7% have actually deployed it across most of their certificates, up 2 % points in a year. And 84% think at least some of their encrypted data is already vulnerable to harvest-now-decrypt-later. They know. They’re just not doing it.

• Anchorage Digital, the federally chartered crypto bank, published what it’s actually running on July 29: quantum-resistant TLS across internal systems and workforce devices, post-quantum algorithms live in its HSMs, and an architecture built so algorithms swap by firmware update. It also open-sourced a Rust implementation of SQIsign. That’s what deployed actually looks like.

H — How It Works: harvest now, decrypt later

The story that actually reaches your systems this fortnight has nothing to do with AI or broken algos.

Picture a room where somebody’s been recording every conversation you’ve had for the last 4 years. They can’t understand a word of it. The language isn’t one they speak, there’s no dictionary for it, and nobody can say when one will exist. So they don’t try. They just label the tapes, file ‘em, and pay the storage bill.

Then, one day, one gets published.

Every tape becomes readable. All of them, at once, going back to the very first one. And nothing about the tapes changed. They sat in a box the whole time. The only thing that changed was who could read them.

That’s your encrypted traffic. Somebody copies it off a wire today, when ML-KEM and RSA are still doing their jobs perfectly, and stores it. Copying is passive. Your systems keep working, the logs stay clean, and the disclosure laws stay quiet, because from the outside it looks like an ordinary day. Nobody will ever tell you it happened.

The dictionary, in this case, is a quantum computer big enough to run Shor’s algorithm. Nobody can name the year. Plenty of serious people say the 2030s, and a serious minority argues the engineering never gets there at all.

And the obvious objection’s fair: who’s actually doing this??? Storage costs $$$, and hoovering up traffic you can’t read is a weird hobby. The answer is that nobody can show you a receipt, because there’s no receipt to show. What we can say is that bulk collection of encrypted traffic is a documented practice going back decades, storage has gotten absurdly cheap since then, and the cost of holding onto a copy rounds to zero. This works without a villain in it. It only takes somebody who noticed that copying a stream costs almost nothing and might pay off enormously, which is just arithmetic.

Your encryption works today, and it’ll still work tomorrow. The number that decides whether that’s enough is a different one: how many years does the thing you sent this morning need to stay secret? Medical records? Decades. Source code and IP? Decades. Attorney-client anything, M&A diligence, the genome you spat into a tube for? Forever, basically.

Run that math and the arithmetic turns on you fast. If your data needs 10 years of secrecy and the dictionary shows up in 8, then the deadline for protecting today’s traffic was 2 years ago. You can’t go back and re-encrypt a conversation somebody already recorded. The tape’s in the box. That decision got made the moment you hit send.

Which is why 84% of those DigiCert respondents saying “yeah, some of our data’s probably already sitting in somebody’s box” while 7% have actually deployed anything is such a WILD gap. Everybody already knows. The standards shipped in August 2024, the browsers finished between 2023 and 2025, and for ordinary web traffic the work is mostly configuration.

Go find out what your own traffic is negotiating. Not next quarter. The tapes are being filed either way.

I — Investment Watch

• Nine firms including BlackRock, Coinbase, Strategy, Fidelity Digital Assets and Anchorage Digital put $15 million over 3 years into a new Bitcoin Security Consortium on July 23, with preparing Bitcoin for quantum computing named among the focus areas. And the group says flatly it won’t direct Bitcoin development or take positions on protocol changes. Worth watching how long that lasts.

• Otherwise a quiet 2 weeks for money. Everyone’s on summer vacation.

F — Field Notes

Every time somebody reads one of these and wants to know what it means for their own systems, the answer used to end in the same place: book a call. I wasn’t being cagey about it. The material genuinely lived in my head and in client documents, and there was nowhere to point them.

That bugged me for about a year. Because most of what people need is just the map. What breaks and what survives, which deadline actually applies to you, what a vendor’s “PQC-ready ✅” means and doesn’t, which numbers in the news are real and which qualifier got dropped on the way to the headline. It’s just scattered across a stack of NIST standards, a dozen national roadmaps, and a pile of papers nobody outside this field reads for fun.

So I built the whole thing and put it online for free. It’s called the Post-Quantum Field Guide, it’s live now, and there’s no email gate, signup, nor tracking. It’s hundreds of pages, every figure traced to its original document, with the wording that states it accurately sitting right next to the wording that gets it wrong.

There’s a Plain English layer for when you need to hand something to a board member or a journalist or your mom, translated rather than dumbed down: same facts, same citations, different vocabulary. And a press section with the charts free to republish.

The reason I’m telling you in this issue specifically: the HAWK coverage last week was better than I expected. A bunch of outlets got it exactly right, and one of ‘em ran with “AI Broke a NIST Candidate. Not Your Encryption,” which is the entire story in eight words. Credit where it’s due.

The ones that slipped, slipped on precision. A few called HAWK a cipher. It’s a signature scheme, which does a completely different job. A few dropped the word candidate, which is the word carrying the whole thing. Neither of those is laziness. Both are just knowing a distinction you had no particular reason to know, and that’s exactly the sort of thing a map is for.

Now there’s somewhere to look. Go break it and tell me what’s wrong with it.

T — The Tangent: 60 hours, $100,000, and one dead algorithm

2022. NIST announces the first winners of its main post-quantum competition and opens a second, smaller one, because almost everything that won was built on the same kind of math. Putting the whole world’s signatures on one mathematical bet is how you end up with one bad afternoon taking down everything at once. So: a side competition, deliberately hunting designs built on different math.

2023 through May 2026. 40 submissions come in. 14 survive to a second round. 9 survive to a third. One of those 9 is HAWK, and HAWK is the only lattice-based one left standing, which makes it the entry that has to justify itself hardest, because lattices are what the existing standards already use. Its pitch was speed and small size. Cryptographers poke at it for years and it holds.

July 28, 2026, late morning. A post goes up on NIST’s public mailing list showing a faster way to recover HAWK’s secret key. The technical content is that HAWK’s underlying math contains a hidden symmetry, and researchers had already worked out that if such a symmetry existed, the scheme was attackable. Nobody had found one. This one was found by an AI model, running about 60 hours, at roughly $100,000 in API costs.

That same afternoon. A human cryptographer replies: “It checks out independently for me.” Just under 5 hours from claim to confirmation.

July 29. HAWK’s own team withdraws it. NIST updates the page the same day. The scheme’s dead, and the side competition drops to 8 candidates with zero lattice options left among them.

So which is it? A standards process working exactly as designed, or a standards process getting outrun?

The case for working-as-designed is strong and it’s the one I’d lead with. This is literally the point of running these things in public for years. A flaw got found, published openly, independently confirmed by a stranger in an afternoon, and acted on by the submitters themselves within a day, before a single product shipped it. Compare that to how you found out about the last vulnerability in something you actually deploy. Nobody got burned because HAWK was still in the lab, and the lab is EXACTLY where you want your algorithms to die.

The case on the other side is about the price tag. $100,000 and 60 hours is inside the budget of a mid-sized company, a university department, or roughly any intelligence service on earth. The multi-year specialist research program used to be the rate limiter on how fast candidates got attacked, and it isn’t anymore. That’s genuinely good for a competition, where more attacks means better survivors. It’s less obviously good everywhere else, and the same write-up reports an improved attack on a weakened 7-round version of AES-128 too. Real AES-128 runs 10 rounds, and AES-256 runs 14. Both fine. But the direction of travel is the thing.

The other half of that worry has an answer, though, and it’s worth saying: whoever finds the flaw, the flaw was always there. HAWK’s weakness existed on the day it was submitted. Everyone who reviewed it across two rounds and missed it left it sitting in a scheme that might have been standardized and shipped into products first. Cheaper attacks mean earlier discovery, and earlier is unambiguously better when the alternative is finding out afterward.

My read: the process won this round, decisively, and the thing to actually take from it is which algorithms you trust. HAWK cleared two rounds of review and folded anyway. ML-KEM and ML-DSA got beaten on in the open for roughly 8 years before anyone standardized them.

Boring, thoroughly attacked, and old is now a feature. Deploy the boring ones.

See ya in two weeks!

Addie LaMarr

Founder, LaMarr Labs

⚛️

Issue #17 | July 22, 2026

(The headliner’s down in Field Notes again this week.)

S — Status Report (The Intelligence)

• AT&T Business and Palo Alto Networks just wired post-quantum crypto into a SASE fabric you can actually buy, folding Prisma into AT&T’s Dynamic Defense with hybrid key exchange and a hardware root of trust. The target is harvest-now-decrypt-later, the game where somebody copies your encrypted traffic today and sits on it until a quantum computer can open it. AT&T’s own word for the motivation: “an operational mandate.”

• IBM got its Nighthawk r2 chip clearing its qubits for reuse 25 times faster, with an 11× drop in start-up errors, and shipped Qiskit v2.5 alongside it. Reset speed sounds like the most boring spec on the sheet. It sets how much work a machine finishes before its fragile quantum state falls apart, which is the number that decides when any of this gets dangerous.

• And Quantinuum, working with UChicago, Harvard and Stony Brook, published a result in Nature that skips one of the ugliest bottlenecks in the field. They got a complete set of quantum operations out of braiding exotic particles around each other on a 54-qubit chip, rather than running the enormously expensive purification step everyone else needs to make operations reliable. Strange, early, and pointed at exactly what stands between today’s machines and useful ones.

H — How It Works: How Two Machines Decide Which Crypto They’re Using

Every secure connection you make starts with a negotiation, and almost nobody who isn’t a protocol nerd has ever looked at it.

Picture two polyglots meeting for the first time with no shared language handed to them in advance. One of ‘em pulls out a list: here’s everything I can speak, in the order I’d prefer. The other one reads down the list, finds the first one she also speaks, and says that one. From that second forward, the entire conversation happens in the language she picked. Nobody renegotiates halfway through dinner.

That’s the handshake. Your browser opens with a list of key-exchange options it supports. The server reads the list, picks one it also supports, and the whole session runs on that choice.

So what changed? The list got a new entry.

Modern browsers now put a hybrid option on the list, usually one called X25519MLKEM768 (try saying that 5x fast), which pairs the classical key exchange everyone’s used for a decade with ML-KEM, the new quantum-safe standard. Belt and suspenders. If the server on the other end also knows that one, you get a quantum-safe connection. If it doesn’t, the two of them fall down the list, land on a classical option they both speak, and carry on perfectly happily. Now it’s an encrypted, working green padlock with zero quantum resistance.

And this is exactly why AT&T and Palo Alto can sell you a “quantum-resilient fabric” as a software change. There’s no rip-and-replace. Somebody adds an entry to a list on a few million servers, and the negotiation does the rest.

There’s a bill attached, though. Picking the quantum-safe option means the server does meaningfully more math every single time a stranger says hello, and it does it before it knows whether the stranger is a customer or a problem.

Hold that thought for about three sentences…

I — Investment Watch

• pQCee raised US$3.9 million in a seed round co-led by SGInnovate and Lotus One, to push its quantum-safe security products into more markets out of Singapore. It’s a small round, but it’s money going into the migration side of the house rather than the hardware side, which is still where the thinner funding lives.

• Quiet fortnight otherwiseeee.

F — Field Notes

TLS handshake exhaustion is an old and frankly boring denial-of-service (DoS) trick. You open a pile of connections, let the server start the expensive part of each handshake, and walk away before finishing any of them. The server burns real work on strangers who were never gonna buy anything. It’s been in the textbooks for twenty-odd years, and it mostly stopped mattering, because classical handshakes got cheap and servers got fast enough to eat the flood and go back to sleep.

A paper last week ran that same attack at a post-quantum TLS server. The old trick works again. Ta-Da! The server sat pinned at 100% CPU for about 25 minutes, and it stayed elevated between attack waves instead of recovering. The identical attack against classical TLS produced a 17-second bump the box shrugged off and forgot about.

Worth keeping the caveat attached: that’s one unhardened server with no rate limiting in front of it, which is nobody’s actual edge. The shape is the finding here, not the size of the number.

The second finding is the one people will get loud about. They pointed two machine-learning intrusion detection systems (IDS) at the attack. One caught about……… half of it. The other scored 0.49 on the metric that runs from 1.0 for a perfect detector down to 0.5 for a coin flip, so, a coin flip with extra steps.

~important note: they ONLY used straight out-of-the-box settings for the IDS~

Yeah. Obviously that’s the result.

Ask anyone who’s actually operated an IDS. Encrypted traffic is a wall, handshake-layer abuse is not something these boxes catch out of the box, and nobody wrote the rules for it. The public datasets these models train on barely contain TLS-layer attacks at all. And here’s the tell that quantum has nothing to do with it: both detectors flopped exactly as hard against ordinary classical TLS in the very same experiment.

So the interesting part sits underneath that headline… Those rules never got written because the attack stopped being worth writing rules for. A dead attack doesn’t earn a signature, a training set, or a dashboard panel. Post-quantum crypto’s what brings it back, and it comes back into a detection layer that quietly retired it years ago.

And, fun fact, we can date the retirement. Back in 2011 a crew called THC shipped a tool that did exactly this at the TLS layer, and the matching bug, CVE-2011-1473, described a client forcing renegotiation over and over inside a single connection purely to burn server CPU. Pull that CVE up today and it's still flagged DISPUTED, on the reasoning that limiting renegotiation is the server operator's job rather than the crypto library's. Which was a defensible call. ‘Twas also the moment the industry filed handshake asymmetry under somebody else's config problem and stopped building detection for it. Post-quantum crypto just reopened the asymmetry, and it's walking straight into the gap that decision left behind.

Which gives you the actual instruction. You can’t slot PQC-TLS into a network and call the migration finished. Switching on ML-KEM and ML-DSA, the new quantum-safe standards for key exchange and signatures, changes the cost of every hello, the shape of your connection backlog, and the traffic profile that every threshold you own was tuned against. Rate limits, connection caps, capacity headroom, IDS and IPS rules, and monitoring baselines all have to move in the same change window as the crypto.

Try this on one service you’ve already migrated: find its connection-rate limit and check when that number was last touched. If it predates the migration, it was set for a protocol you’re no longer running.

The algorithm swap is the easy part, and it’s the only part most migration plans have a line item for.

T — The Tangent: The Algorithm Nobody Picked Just Became the First One ISO Blessed

1978. A Caltech engineer named Robert McEliece publishes a public-key cryptosystem built on error-correcting codes, the same branch of math that lets a scratched CD still play. It’s clever, it’s fast, and it has one glaring flaw for 1978: the public keys are enormous. Everyone admires it and nobody ships it, because RSA arrives around the same time with keys you can print on a business card. McEliece goes in the drawer.

The next 4 decades. Cryptographers keep poking at it, the way you’d keep rattling a doorknob. Attack after attack gets thrown at it, and it holds. Every one of them. Meanwhile RSA and elliptic curve go on to encrypt basically the entire internet, right up until Shor’s algorithm arrives, the quantum recipe that unpicks the math both of them rest on, and quietly guarantees a big enough quantum computer eventually takes them apart.

2016 onward. NIST opens its post-quantum competition and Classic McEliece is right there in the field, now maintained by an international team. It has the best security story of anyone in the room: 40-plus years of failed attacks and not a scratch. And it loses. NIST goes with ML-KEM, built on lattices, a completely different hard math problem that produces pleasantly small keys. McEliece’s public keys start around 261 kilobytes and climb past a megabyte at the settings people actually recommend for long-term security, and nobody wants to ship a small novel just to open a web connection. Even later, when NIST goes shopping specifically for a backup built on error-correcting codes, it picks HQC instead.

July 15, 2026. ISO adds Classic McEliece to its asymmetric cipher standard, ISO/IEC 18033-2, making the 1978 design the first post-quantum algorithm ISO has ever standardized. Germany’s BSI and the Dutch NCSC were already recommending it. Now organizations across ISO’s 177 member states have an international standard to point at.

So which is it, vindication or a glorified museum piece?

The case for vindication is the strongest one in cryptography: time. Nothing else in the post-quantum field has stood in the open for 48 years and survived. The lattice designs are excellent and they’re also young, and 2026 has already produced a couple of scares about breaking them. Having a standardized fallback built on completely different math, so a break in one family doesn’t take the other down, is exactly the insurance policy a serious government cares about. Which is precisely why the German and Dutch agencies got there first.

The case against is simpler and harder to argue with. Standards don’t deploy anything. Those keys are still enormous, TLS still won’t have them, browsers still won’t ship them, and the practical footprint stays where it’s always been: high-value, low-volume, long-life links where you can afford the bandwidth and you need the secret to survive for thirty years. That’s a real and important niche. It’s also a niche.

My read is that both are true and the interesting bit is who moved. ISO standardized the algorithm the market passed over, right as the rest of the world locks onto a single lattice family for everything. Somebody wanted a second door, and they weren’t willing to wait for NIST to want one too.

See ya in two weeks!

Addie LaMarr

Founder, LaMarr Labs

⚛️

(The headliner’s in Field Notes this week.)

S — Status Report (The Intelligence)

• Microsoft just pulled its own quantum-safe deadline up to 2029, a full year ahead of the federal government’s 2030 line, and said the quiet part out loud: the risk timeline’s moving faster than everyone’s roadmaps assumed. When the company half your enterprise stack runs on decides it can’t wait, that’s the tell worth reading.

• Over in Andhra Pradesh, IBM is spinning up one of India’s first real quantum computers, a 156-qubit System Two it plans to switch on by September. That’s another country, another machine, and another few hundred qubits that didn’t exist last quarter. The hardware curve dgaf about your migration backlog.

• And the Department of War rolled out its own post-quantum strategy, a 25-page mandate ordering every military system, from battlefield radios up to nuclear command-and-control, onto quantum-safe crypto by the end of 2031, with anything that misses the date slated to get retired. When the people guarding the nukes put a hard clock on it, “someday” stops being a strategy. (ngl I HATED calling it the DOW)

H — How It Works: Why Your New Quantum-Proof Signature Is So Freaking Fat

I’ll give you a quick setup, then the picture that makes it click.

Every time your browser confirms a website is really that website, it checks a digital signature. For decades those signatures have been teeny tiny. An old-school ECDSA signature runs about 64 bytes. That’s smaller than this sentence.

The new quantum-safe signatures are BIG. Comically big.

Now let’s put that ~in human terms~.

Picture the mail slot in your front door. Like the skinny little brass flap, built for envelopes. For years that’s exactly what you’ve been getting through it: envelopes. Thin, flat, slide right through, donezo. That’s your ECDSA signature. 64 bytes of paper going through a slot built for paper.

Then the quantum-safe upgrade shows up and hands you a mattress.

ML-DSA, the main new signature standard, lands somewhere around 2,400 to 4,600 bytes. That’s a fat parcel you’re now trying to cram through the flap. And the extra-paranoid option, the hash-based SLH-DSA… it’s up to about 50,000 bytes. That’s the king mattress. You’re on the porch at midnight folding a mattress through a brass mail slot, and the slot is every protocol on the internet that got designed back when signatures were small.

So what tears?

Handshakes bloat, because your browser and the server now sling these fat things back and forth every single time they say hello.

Certificates balloon. And anything living on a skinny pipe, an IoT sensor, a car, a 5G tower broadcasting to a whole city, suddenly has to chop the mattress into a dozen pieces and pray they all show up. Some of these systems just can’t. The slot’s too small and there’s no landlord to call.

And why does this matter right this second? That shiny executive order from last month put the signature migration on a hard 2031 federal clock. So a whole lot of teams are about to find out, live, that the quantum-safe upgrade is a fat mattress, and their front door was built for envelopes.

So when a vendor swears they’re “quantum-ready,” the (not so) fun follow-up is: ready….? how, exactly? Getting the math right was the easy part. Getting the mattress through the slot is another story entirely.

I — Investment Watch

• IQM finally closed its SPAC and started trading on Nasdaq as “IQMX” on July 2, turning the Finnish quantum-hardware maker into a public company. The catch: once shareholders decided whether to stay in or cash out, the combined company walked away with about $233 million, a good chunk lighter than the deal first dangled. Going public and raising what you hoped for turn out to be very different sports.

• Days earlier, SEEQC filed to go public too, lodging its S-1 on June 29 to list on Nasdaq as SEQC through a merger with a blank-check company, at roughly a $1 billion valuation. Two quantum-hardware names reaching for the public markets inside of a week.

• And the money’s flowing toward the actual fix too. Wultra, a Prague outfit, raised €6.8 million to scale post-quantum authentication for banks bracing for AI-driven identity fraud and the quantum migration at the same time. Small round, telling one: it’s the exact AI-and-quantum collision this issue keeps circling, wearing a VC term sheet.

F — Field Notes

Every security team I know is treating AI and quantum as two separate problems on two separate roadmaps. One’s this year’s fire, one’s a someday. And I get why. But they’re the same war, and once you see it you can’t unsee it.

The war is this: both of them attack your ability to know that anything is real.

AI attacks it from the outside. It manufactures fakes good enough to sail past the humans and systems that used to catch them: cloned voices, pixel-perfect login pages, synthetic faces opening fresh accounts. Quantum attacks the same thing from the inside. It breaks the digital signatures that are a machine’s way of proving something’s genuine, and once Shor’s algorithm cracks that signature, a forgery comes out mathematically identical to the real thing. There’s no tell nor tamper mark, and nothing left to catch it.

It’s the same target, split across two directions. AI counterfeits the thing; quantum forges the seal that was supposed to catch the counterfeit. So there’s really only one battlefield here: the layer where you prove who and what is authentic: your identity and authentication stack. You harden it to be hard to fool (that’s the AI fight) and hard to forge (that’s the quantum fight), and that’s one system getting one overhaul. Two separate programs would just run right past each other.

And if you still think these are two different fires, look at what’s already happening. Researchers are using machine learning, the exact AI everyone’s funding, to pull secret keys straight off the chips running the new quantum-safe algorithms, reading faint power and timing leaks no human could catch. The attacker’s toolkit doesn’t care about your org chart. The AI is already aimed at the quantum defense.

None of this collapses two specialties into one hire. You still need your crypto people and your AI people. Just point them at one thing, the trust layer, instead of two, and you’re fighting one war with one plan.

T — The Tangent: Why the Government Gave Signatures an Extra Year (Is That Genius or a Cop-Out??)

There’s a detail buried in that executive order almost nobody stopped to clock: the two deadlines aren’t the same. Federal agencies have to switch their key establishment, the encryption that hides your data while it’s moving, to quantum-safe by the end of 2030. But their digital signatures, the stuff that proves who’s actually who, get a whole extra year, until the end of 2031.

One year of daylight between two halves of the same migration. Looks like a rounding error. It’s actually a real decision, and there’s a genuine argument on both sides.

The case that it’s smart. The two jobs carry totally different risk. Encryption is exposed to harvest-now-decrypt-later, that patient-adversary game where someone copies your encrypted traffic today and just sits on it, waiting for a quantum computer to catch up and pop it open in 2032. Anything you send right now that has to stay secret for a decade is already at risk, today, before a single big quantum computer exists. So encryption is genuinely the fire to put out first.

Signatures don’t work that way. A signature only has to hold up at the exact moment it gets checked. Nobody can reach back and forge a 2026 signature in 2032 to rewrite history. So the day the quantum computer shows up, you need your signatures quantum-safe from that point forward, but there’s no backlog of old traffic quietly rotting. Doing encryption first and signatures second is triage, plain and simple: you treat the wound that’s already bleeding before the one that hasn’t started. And the practical kicker, signatures are the fat mattress from earlier, the harder, bulkier, more disruptive swap. Sequencing the easy migration ahead of the ugly one is just how you’d run any big project without lighting it on fire.

The case that it’s a cop-out. Now flip it. That extra year is exactly the kind of wiggle room that curdles into an excuse. Everybody in security knows the pattern cold: the thing with the later deadline is the thing that gets ignored until the deadline’s basically tomorrow. Give the harder, uglier migration MORE runway and less urgency, and you’ve all but guaranteed people start it dead last, panic, and botch it. The bulky one needed a head start. Instead it got a hall pass.

There’s a subtler problem too. Splitting the timeline quietly tells everyone that signatures matter less, when for plenty of systems, authentication is the exact thing standing between you and a genuinely terrible day. An org that’s locked down its encryption while its signature infrastructure limps toward some far-off 2031 finish line is an org with a wide-open door for anybody who can forge trust.

So which is it? Honestly? Both. The sequencing logic is sound, the harvest-now math really does make encryption the first fire. And the human-nature read is also right, that extra year is going to get abused by every team looking for permission to procrastinate on the hard part. The order did a defensible thing and handed everyone a ready-made excuse in the very same sentence.

The tell’ll be simple. The orgs that read 2031 as “start now, it’s the harder one” are going to glide. The ones that read it as “cool, later” become the case studies. Same deadline, opposite endings, and you already know today which kind of shop you’re running.

See ya in two weeks!

Addie LaMarr

Founder, LaMarr Labs

(The actual argument’s in Field Notes this week.)

S — Status Report (The Intelligence)

• On June 22 the White House signed an executive order with the friendliest possible name, “Securing the Nation Against Advanced Cryptographic Attacks,” that quietly put a hard clock on the whole federal government’s switch to post-quantum crypto. And it reaches past the agencies: it tells the people who write federal contracting rules to drag contractors onto the same timeline, and it landed next to a second order standing up a federal quantum-computer buildout. So if you sell to Uncle Sam, your migration just got a due date. The exact dates, and why they’re the easy part, live in Field Notes.

• A new measurement study crawled 32,011 domains to see what the internet’s actually negotiating, and found 15.7% still riding TLS 1.2, heaviest in (of all places) banking and government. The exact two sectors that order is yelling at. Deadlines on the calendar, 2019 on the wire.

• Meanwhile researchers ran Regev’s algorithm, a newer cousin of Shor’s (the math that eventually cracks RSA), on real quantum hardware and lined it up against Shor’s head to head. The number this fancy quantum machine managed to factor? 15. As in three times five, the thing a ten-year-old does on a napkin. The threat’s a “harvest your data now, crack it years later” problem, not a this-quarter one.

H — How It Works: Why There Are TWO Post-Quantum Algorithms

When NIST finished its post-quantum bake-off, it didn’t crown one winner for protecting your key exchange. It standardized ML-KEM, built on a flavor of math called lattices, and then a few months later it picked a second one called HQC, built on completely unrelated math (error-correcting codes). Same job. Two totally different ways of doing it. On purpose.

Now let’s say that ~without the standards-committee voice~.

Picture two locks on your front door. Not two of the same deadbolt, that’d be pointless. One’s a pin-tumbler lock, the kind a burglar pops with a little rake and a tension wrench. The other’s a magnetic combination, with no keyhole, zero pins, and nothing to rake. It’s got totally different guts.

Now imagine a burglar spends TEN YEARS becoming the single greatest pin-tumbler picker who has ever lived. Walks up to your door, beats the first lock in four seconds flat… and then just stands there. Because every hour he poured into pins taught him exactly nothing about the magnet. The skill doesn’t transfer. He’s back to zero on lock number two.

That’s the whole idea behind picking HQC. ML-KEM is the pin-tumbler. HQC is the magnet. If some brilliant mathematician (or some big quantum machine) ever figures out how to pop lattice math, that breakthrough cracks ML-KEM wide open… and does basically nothing to HQC, because HQC isn’t made of lattices. It’s a spare built out of different bones.

Here’s the part people miss. A backup that runs the SAME math as the primary is a big, fat placebo. If lattices fall and your “backup” is also lattice-based, you didn’t hedge a thing. The entire point of HQC is that it shares zero DNA with ML-KEM. One breakthrough can’t take both out.

It’s slower and the keys are chunkier, which is exactly why nobody’s rushing to deploy it as the daily driver. That’s fine. You don’t keep a spare tire because it’s a great tire. You keep it because the day you need it, you REALLY need it. So when a vendor swears they’re “post-quantum ready,” the follow-up question is which algorithms, and whether they can actually run the spare if the main one goes flat.

I — Investment Watch

• IQM’s shareholders vote June 25 on a SPAC merger that, if it clears, lists the Finnish quantum-hardware maker on Nasdaq as “IQMX” and makes it the first European quantum computing company on a major US exchange. It was pitched at a $1.8 billion valuation with up to ~$465 million of cash earmarked for fault-tolerant hardware, though SPAC redemptions can shave that down before it closes. The same machines that EO’s 2030 clock is quietly betting either will or won’t show up in time.

• Worth clocking where that money’s flowing. Crunchbase says private quantum VC is actually down in 2026, with the real action sliding to the public markets, so IPOs and SPACs are doing the heavy lifting now. Translation: the big names that already raised are cashing out to retail while the early-stage faucet runs thinner. And whose bag that becomes? Depends entirely on which side of the listing you’re standing on.

F — Field Notes

Here’s what that executive order actually puts on the calendar. Federal civilian agencies have to move their high-value systems to post-quantum key establishment (the handshake that protects data in transit) by the end of 2030, and their digital signatures by the end of 2031. NIST owes a pilot migration by the end of 2027. And the Federal Acquisition Regulatory Council, the folks who write the rules every contractor lives by, got told to draft a rule dragging contractors onto that same 2030 line. So if you sell to a federal agency, this is your timeline too.

And in every migration conversation that’s about to kick off because of it, the first reaction’s going to be identical. “Okay, 2030, we’ve gOt TiMe, throw it on the roadmap.” That reaction is wrong, and not by a little.

A deadline quietly assumes one thing: that you know how long the move takes. Most shops have no clue. So the question this EO should actually make you ask is your rotation time. Forget “are we post-quantum ready,” which is a yes/no you can ~comfortably~ lie to yourself about in a status meeting. Rotation time is the honest one: if the algorithm you’re running today got broken tomorrow morning, how many days, weeks, or quarters until the replacement is actually live EVERYWHERE it needs to be??? There’s even a paper out this month proposing rotation time as a real, measurable number, and tying how much you can tolerate straight to how much risk you’re carrying.

For most orgs that number is measured in years, and almost nobody’s ever actually measured it. Here’s the cheap gut-check: pick ONE system and ask the team how fast they could swap its key-exchange algorithm if they were forced to this week. Watch the answer go vague real quick. Now take that vagueness and multiply it across every service, vendor endpoint, and load balancer you don’t directly control. THAT number, not a green checkbox on a compliance slide, is the thing actually standing between you and a 2030 you can hit. Getting it measured and shrunk across a whole estate is the real work, and it’s the work a dashboard will keep swearing you’ve already done.

T — The Tangent: NIST already pulled this exact move once, and the spare’s been sitting in a drawer for 13 years

Picking a second algorithm built on different math isn’t NIST improvising under quantum panic. They ran this precise play almost two decades ago, and here’s the beautiful part: almost nobody ever needed the result. Which is the whole point of doing it.

2007. NIST opens a public competition for a brand-new hash function, the thing that turns a file into a unique fingerprint, to be called SHA-3. And the weird thing about this competition? The hash function everyone was already using, SHA-2, was totally fine. Not broken. No emergency. NIST ran the contest anyway, on a hunch, as a just-in-case. 64 submissions poured in from cryptographers all over the planet, eventually whittled down to five finalists with very metal-band names: BLAKE, Grøstl, JH, Keccak, and Skein.

2012. Keccak wins. And the reason it won is the entire moral of this issue. Keccak was built on a design called a “sponge construction,” which works nothing like SHA-2’s internal machinery. NIST picked it SPECIFICALLY because it was so different, so that if SHA-2 ever did fall, SHA-3 wouldn’t tumble down with it. Bruce Schneier flat-out called it an insurance policy the day it was announced.

And then? Nothing happened. SHA-2 still hasn’t broken, 13 years later. SHA-3 has sat in the standards drawer that whole time as a spare nobody’s had to grab. That’s not a swing and a miss. That is EXACTLY what a good hedge looks like when the thing you were scared of politely declines to happen. You pay a small premium, the disaster skips you, and the premium feels like a bargain in hindsight.

HQC is the same bet, re-run for the quantum era. ML-KEM (the lattice one) is the SHA-2 of this story, the workhorse everybody’s actually deploying. HQC (the code-based one) is the Keccak, the spare built on different bones and standardized now, on a hunch, so that if lattice math ever cracks there’s already a swap waiting that doesn’t share the flaw. And that fresh EO deadline is the thing that turns the spare from a nice idea into something you’d genuinely better be able to reach for in a hurry.

Which loops us right back to your rotation time. A spare key’s only worth having if you can actually get to it before the door’s already open. NIST did the hard part and built you the second lock. Whether you can swap to it in time is the part nobody can do for you.

See ya in two weeks!

Addie LaMarr

Founder, LaMarr Labs

⚛️

Issue Lucky #13 | June 10, 2026

Headline: The choir part is in the How it Works & Tangent sections

S — Status Report (The Intelligence)

• Microsoft hopped onstage at Build and bragged about a “1000x” reliability jump on its shiny new Majorana 2 chip. Then a physicist at St Andrews read the same paper and basically went “this fixes none of the actual problems”… so maybe keep it off the board deck for now.

• IQM dropped a new flavor of error-correcting code it says needs up to 8x fewer of the junky little qubits to babysit one good one. THAT ratio is the number that matters, not the big qubit count everyone loves to brag about.

• And QuiX bolted on a real-time “feed-forward” gadget that fixes a photonic computer’s screwups in about 150 nanoseconds, before the next photon even shows up. Three totally different machines, same two weeks, all chasing the exact same thing: keeping the math from falling apart mid-calculation.

H — How It Works: Quantum Error Correction

Quantum error correction is how you stop a quantum computer from face-planting halfway through the math. Qubits are ridiculously fragile lil things, they catch errors from basically everything (a stray vibration, a warm photon, you breathing near them, practically), and a long calculation turns to mush unless something’s cleaning up the whole way through.

Now let’s say that ~without the lab coat~.

Picture a choir. A hundred people strong, all holding the same note. One of ‘em drifts flat. You don’t need to sprint around checking everybody’s sheet music to find the culprit, right??? You just HEAR it. The note goes sour against the group, the singers next to them lean back on pitch, and the sound stays clean even though not one person in that room is actually perfect.

That’s the whole trick. The note is your precious little piece of quantum info. The hundred singers are the physical qubits holding it up. The sour wobble’s an error. The lean-back is the fix.

And here’s the genuinely weird part. You can’t just LOOK at a qubit to check on it. Staring at a quantum state wrecks it, the same way grabbing one singer and going “hey, solo that real quick” instantly kills the harmony. So error correction never checks the singers one at a time. It listens to the whole GROUP, hears the disagreement, and fixes the flat note without ever once reading your actual secret. Catches the mistake, never sees the answer. Wild.

Every code in the news this week is just a different seating chart for the choir. The surface code, IQM’s new barbell codes, all of ‘em are rules for how many singers you need and where to sit them so one flat voice can’t tank the whole song. “8x fewer qubits” is a smaller choir that still holds pitch. Majorana’s bet is singers who just don’t go flat as easy. Same goal, opposite ends.

So when you see a headline screaming about some giant qubit count, that’s just the size of the choir. Cool. Big choir. The number that actually tells you whether the song survives to the last note is the error rate AFTER the choir does its thing, the “logical” one, and THAT’s the number that says how close anybody really is to running the math that cracks RSA. (Funny how it never makes the headline. Way less sexy than a big round number.)

I — Investment Watch

• OQC closed a £260 million (~$350M) Series C, the biggest private quantum round Europe’s ever pulled off, and the cash is going straight into fault-tolerant hardware, aka machines built to not fall apart mid-calculation. Same exact thing the whole news cycle’s chasing this week.

• Buuut one fat round isn’t a boom. Crunchbase says 2026 quantum funding is actually DOWN from last year ($1.2B vs $4.1B), with something like 92% of it crammed into a tiny handful of mega-rounds, so “quantum funding is booming!!” is a headline doing a lot of quiet lifting.

F — Field Notes

In basically every TLS config, somebody flipped on TLS 1.3 ages ago, checked the box, and walked away. Scan says 1.3. Box is green. Done!

Except the config underneath is STILL cheerfully accepting TLS 1.2, sometimes older, with plain classical key exchange, because nobody ever went back and switched the geezer stuff off.

A handshake shows you the fanciest thing your server CAN do. It says nada about the worst thing it’ll still say yes to. One picky client, or one attacker who quietly strips the good option out of the negotiation, and your shiny 1.3 connection slides right back down to 2019.

Now the part that actually matters for quantum: TLS 1.3 is NOT post-quantum. The TLS version and the key-exchange group are two totally different knobs, and people smush ‘em together constantly. You can run a flawless 1.3 handshake that’s still doing plain old X25519 with zero post-quantum anything in it. The thing you actually want on the wire in 2026 is the hybrid group, X25519MLKEM768. So a “PqC sUpPoRtEd ✅” line in some doc only means the box CAN do it. Tells you nothing about whether it’s actually DOING it.

(There’s even a paper out this month making this exact point with a straight face: the math’s been ready for ages, the migration’s stuck because the boring config grind never gets handed to an actual human. Your TLS config is that grind, in the flesh.)

There are so many configs that say 1.3 on the tin and still happily talk 2019 underneath. So here’s the cheap version: that green checkmark is bragging about what your server’s proud of, not what it’ll settle for. Go poke ONE endpoint, see what it actually still accepts, and it’ll rattle you enough to stop trusting the dashboard. The real monster isn’t that one host though. It’s the hundreds of services, load balancers, and inherited vendor endpoints quietly terminating TLS in ways nobody’s looked at since the day they were stood up, where “we turned on 1.3” and “we turned off the old junk” are secretly two completely different facts. THAT gap, across a whole estate, is the actual migration. And it’s the exact thing a dashboard will keep swearing up and down is green.

T — The Tangent: the 1952 lectures that already called all of this

None of this error-correction hullabaloo is new, btw. The core idea is like 70 years old, and it came from a guy trying to build a working computer out of parts that kept dying on him.

It was 1952. John von Neumann gives a run of lectures at Caltech, later written up in 1956 under the greatest title in all of computing: “Probabilistic Logics and the Synthesis of Reliable Organisms from Unreliable Components.” (Try saying that at a party.) The machines back then ran on thousands of vacuum tubes, and tubes burned out CONSTANTLY, so the computer was basically broken for a chunk of every single day. Everybody assumed the obvious thing: chain a bunch of unreliable junk together and it only gets worse as you add links.

von Neumann flipped it. What if you pile on redundancy on purpose? Run the same logic through a bunch of copies, take a majority vote, and you can drag the error rate of the whole machine DOWN even though every individual part stays trash. He did the actual math: below a certain per-part error threshold, you can make the whole thing as reliable as you want, just by throwing in enough redundancy. That’s the choir. Proven on paper. A trustworthy machine built out of singers who keep going flat.

And that’s almost exactly the bet every quantum error-correction scheme is making right now. The “threshold theorem” sitting under all of fault-tolerant quantum computing, the whole reason IQM’s barbell codes and the surface code and Microsoft’s chip are even worth funding, is the great-grandkid of what von Neumann scribbled before the transistor was even a year old. The qubits are the dying tubes. The codes are the redundancy. The threshold is the line where the trick switches on.

Oh, and he pulled this off while ALSO inventing the computer architecture sitting in your laptop right now, working on the H-bomb, and casually founding game theory on the side. The reliable-from-unreliable paper is basically a footnote in his career. It just happens to be the footnote that 2026’s most expensive hardware is sprinting to physically build.

So all three of those machines burning the same two weeks on the same problem? They were really just catching up to a 1952 idea, and the hardware finally crawled close enough to the threshold for the old math to start paying out. Q-Day, the day a machine can finally read your encrypted traffic, is just the day somebody’s choir holds the note long enough to make it to the end of the song.

See ya in two weeks!

Addie LaMarr

Founder, LaMarr Labs

Issue #12 | May 26, 2026

(The headliner’s in the Field Notes section this week)

S — Status Report (The Intelligence)

• Three cryptographers tried to break lattice crypto again this week. They posted a paper on May 19 claiming a faster quantum attack on the math under ML-KEM (the new NIST encryption standard your PQC migration plan is built on). Did it work??? Nope. The fast version of their attack needs a quantum computer with more memory than physically exists. The slow version exists but runs… slowly. The authors themselves basically said: “these speedups may be difficult to ever realize in practice.” Translation: cool paper, no actual break. Standards intact. The juicy drama resides in the Tangent section.

• NIST just promoted nine more post-quantum signature candidates to Round 3 on May 14, which is the standards process quietly hedging against the swing-and-miss in bullet one. Four math families on the list: lattice (HAWK), MPC-in-the-Head (FAEST, MQOM, SDitH), multivariate (MAYO, QR-UOV, SNOVA, UOV), and isogeny (SQIsign). If lattice ever does break, the standards process has been seeding alternates for three years. Round 3 runs about two years. Tweak submissions due August 14.

• The White House is circulating a draft executive order that does what Congress hasn’t gotten around to. Federal contractors get a 2030 PQC deadline. Digital-signature transitions on high-impact systems get December 31, 2031. Mechanism: procurement, not statute. Which means → if you sell to a covered agency, your PQC roadmap just got a contractual due date even if the law never shows up.

H — How It Works: Lattice Crypto

Lattice cryptography is the math that hides your secrets inside a problem so geometrically twisted that even a quantum computer stalls out on it. ML-KEM is the most important post-quantum standard built on it, and it’s the algorithm your migration plan is mostly about.

Now let’s translate that ~out of nerd~.

Imagine you and your best friend agree on a meeting spot in a normal city. You name a street, a building, and a floor. Anyone walking around with a map can find you.

Now imagine the city has 500 different layers stacked on top of each other. Each layer’s identical-looking, but offset in some weird specific direction nobody else knows. You’ve got the address. Your friend’s got the address. Everyone else on Earth has to search every possible layer at every possible offset, and the number of spots they have to check is bigger than the number of atoms in the universe.

That’s lattice crypto. Your secret is the meeting spot. The 500-layer city is the math. The address is your private key. And the part where even a quantum computer stalls is the city having 500 layers instead of one. (Shor’s algorithm, the quantum thing that breaks RSA and ECC, is incredible at cracking number locks but has no idea what to do with 500-dimensional geometry. It’s a totally different math problem.)

When you actually send a secret with ML-KEM, what’s really happening: you’re tucking the secret into one specific spot in that 500-layer city and throwing a cloak of random noise over the whole package so nobody can read it. Your friend’s private key knows exactly how to lift the cloak off and find the spot, like noise-cancelling headphones for math. Everyone else just sees fog. And if someone tries to mess with the package on the way, the math falls apart in a way your friend can see… aka tampering basically snitches on itself.

Last piece: Early lattice schemes were beautiful and totally unusable on real hardware (way too slow). The trick that got ML-KEM through standardization is called the NTT, which takes one expensive operation and fans it out into a bunch of cheap independent ones, like a glass prism splitting white light into colors. The prism is why ML-KEM fits inside a TLS handshake without setting your laptop on fire.

This stuff is impossible to visualize, so I used Manim to render the hard math. It’s from my series called Visualizing FIPS 203, which will come out later this year.

So: 500-layer city + cloak of noise + prism speed trick. That’s the three pieces of lattice crypto. Next time your vendor sends you a deck with “PQC supported ✅,” ask them which of those three pieces their wire is actually negotiating. (Spoiler: they have no idea.)

I — Investment Watch

• Quantinuum filed for its US IPO on May 26, targeting a $12.7 billion valuation at $45-50 per share, planning to raise up to $1.05 billion on Nasdaq under “QNT.” Honeywell and Cambridge Quantum keep ~82% of the equity. First real S-1, no SPAC. Which means the SEC ~actually~ reads it. The trading band over the next quarter is the first real public-markets test the entire quantum sector has faced… everyone else hovering around the IPO door is watching this one very carefully.

• Commerce just signed $2 billion across nine quantum companies in CHIPS Act letters of intent, the largest single federal commitment to the sector ever. The list leans hard toward the fault-tolerant track (the long-game machines), skipping the NISQ end (the short-term noisy ones). PsiQuantum’s $100M LOI from May 21 is the biggest individual award. The thing to notice: state capital is becoming a bigger slice of the quantum stack, faster than venture is. And LOIs live in their own category: commitments to negotiate, sitting one step before any real grant or contract gets signed. That ambiguity is the part you should be modeling in your vendor risk over the next two years.

• Project Eleven published a Q-Day threat report putting baseline Q-Day at 2033 (Q-Day = the day a quantum computer can break RSA / ECC for real), with optimistic edge 2030 and pessimistic edge 2042. The interesting part is the years themselves matter less than the fact that the conversation finally shifted from “is Q-Day even real” to “which years does your migration calendar actually key off.” Bitcoin, Ethereum, every other ECC-anchored ledger sits in that 2030-2042 window. If your migration calendar still says “post-2035,” the honest question is who picked 2035 and what they had on paper.

F — Field Notes

In any 2026 PQC migration conversation, when uptake is slow, the default move’s to JuSt BrIeF hArDeR. Feed ‘em moar board memos, sharper deadlines, and spookier breach disclosures. As if lecturing is the missing piece.

Well, the move is wrong by about 40 years.

You can talk yourself hoarse about quantum risk and almost nobody will move. Most of the people who needed to hear it tuned it out, the ones who heard it nod, the deck gets a checkbox, and the wire keeps negotiating X25519 like it’s 2019. So, what actually moves people?!?!

A friggin’ movie.

OK exhibit A.

Ronald Reagan watched WarGames at Camp David the weekend it dropped in June 1983 (for anyone who’s never seen it: teenager hacker accidentally almost starts WWIII from his bedroom modem). Reagan apparently kept going back to it, pulled General Vessey aside at the next NSC meeting, asked basically “can this actually happen?” Vessey came back a week later with: “Mr. President, the problem is much worse than you think.” Fifteen months after that movie night, NSDD-145 landed on Reagan’s desk, signed September 17, 1984: first US national-security directive on telecommunications and information-systems security. Years of classified briefings sat there doing absolutely nothing. Yet, one teenager on one movie screen moved the needle.

Exhibit B.

Mike McConnell, who would later run the NSA, has said in talks that Sneakers in 1992 (the Robert Redford one where the heist crew steals the magic decryption box) sharpened how he framed the cyber mission as he was taking over the agency. Just to be clear: that’s the leader of the agency with access to literally all of the briefings in the world, and the thing that crystallized his framing was… a movie.

Fun fact: WarGames and Sneakers were co-written by the same two screenwriters, Lawrence Lasker and Walter F. Parkes. They’d been trying to (unsuccessfully) warn people about computer-network security since the early 80s, conceived Sneakers during WarGames production, and ended up moving both Reagan’s policy and McConnell’s framing through their stories.

Briefings couldn’t do it. But their movies did.

So when somebody on your team is supposed to be migrating to PQC and has only ever heard the threat talk, stop briefing and start telling stories about who saved the internet of Christmas past.

Tell them about Cliff Stoll, the Berkeley sysadmin who in 1986 noticed a 75-cent accounting error on his system, pulled on the thread, and ended up catching a KGB-linked hacker exfiltrating from US military networks (he wrote it up in The Cuckoo’s Egg and basically taught a generation of incident responders what the small weird signal nobody cares about can turn into). Side note, just finished the book yesterday and it’s a MUST-READ for anyone excited about infosec.

Tell them about Dan Kaminsky in 2008, who found a hole in DNS that could’ve collapsed the internet, then ran a secret patch summit at Microsoft’s Redmond campus with every major vendor on the planet before anyone could weaponize the bug.

Tell them about Whit Diffie and Marty Hellman in 1976, two outsider cryptographers who invented public-key cryptography from scratch and gave the modern internet its locks before industry even asked for them… while also fighting academia.

These are the heroes of the internet of Christmas past. The PQC migration’s the next chapter. The engineer on your team who actually flips the config from X25519 to ML-KEM hybrid are in the lineage.

Make them see it.

The 1983 lesson and the 1992 lesson both come out the same way: people move when they can see themselves inside the story.

T — The Tangent: Lattice Crypto Just Survived Another Assassination Attempt

OK so here’s the actual reason cryptographers are losing it about lattice crypto this week. It goes back two years.

April 2024. A researcher named Yilei Chen at Tsinghua University drops a 60-page preprint claiming he’d found a polynomial-time quantum algorithm to break lattice crypto. Plain English: he was saying the math under ML-KEM, ML-DSA, and basically all of NIST’s post-quantum portfolio was broken. If true, the entire federal migration calendar, OMB memos, CNSA 2.0 timeline, all of it built on sand. Cryptography Twitter loses its mind for about a week.

Eight days later, the bug gets found. Hongxun Wu at Berkeley and Thomas Vidick at the Weizmann Institute, working independently, both spot a fatal mistake in Step 9 of Chen’s algorithm. Chen acknowledges it himself: “Step 9 of the algorithm contains a bug, which I don’t know how to fix.” The paper gets pulled from the spotlight as fast as it had landed.

But the cryptographers can’t agree on what the whole thing actually MEANT. Adi Shamir (the S in RSA, so kind of a big deal) calls the result “psychologically significant” — meaning even if Chen’s attack didn’t work, the next one might. Craig Gentry, the guy who invented fully homomorphic encryption on lattices, basically rolls his eyes… a non-working attack is just an interesting failure. Bruce Schneier lands in the middle: the public standardization process is itself the hedge against any one paper being right. They argue about it for months.

Cut to two years later. May 19, 2026. Chevignard, Shen, and Schrottenloher post the next swing. Way narrower than Chen’s was — they’re claiming a quantum speedup on just one piece of the attack, not the whole thing. The fast version needs more quantum memory than physically exists. And the authors literally say in their own paper: “such asymptotic speedups may be difficult to ever realize in practice.” They wrote a toolbox paper. The headlines are louder than the paper.

That’s the drama. Two years, two swings, two outcomes: standards still standing. What you’re watching is cryptanalysis doing its job in public, with receipts. Q-Day, if it ever shows up, arrives as a working algorithm that survives 8 days of cryptography Twitter. Still waiting on the first one.

See ya in two weeks!

Addie LaMarr

Founder, LaMarr Labs

⚛️

The Quantum S.H.I.F.T.

Issue #11 | May 13, 2026

S — Status Report (The Intelligence)

• IonQ and Florida LambdaRail lit up the first statewide U.S. quantum-safe network: a 100-mile quantum-encrypted corridor from Palm Beach to Miami-Dade on real production fiber, not a press-release pilot.

• The Harvard Quantum Initiative pulled its fault-tolerance timeline forward five to ten years on May 4, with large-scale quantum systems landing inside this decade rather than the 2030s. Every “ten to fifteen years out” migration plan just got a new floor.

• Cloudflare edited its full PQC migration roadmap on May 6: post-quantum origin authentication mid-2026, Merkle Tree Certificates mid-2027, full SASE PQ by early 2028. Over 65% of human traffic through Cloudflare already uses post-quantum key agreement by default.

H — How It Works: Side Channels (or, Why the Math Isn’t the Whole Story)

A side-channel attack pulls the secret key by reading what the implementation never meant to send: how long the math took, how much power it drew, what the chip emitted. The math stays intact. The box around the math leaks like a tea bag.

Now let’s translate that for normies.

Picture a safecracker who can’t see the dial. She listens. Every turn of the wheel, the tumblers click at a slightly different volume depending on whether the digit is part of the combination. Thirty seconds later she walks out with the combination, having never touched the dial.

That’s a side channel. The math behind the lock is perfect. The box around the math tells the attacker the answer through a channel that was never supposed to be there.

This stuff is old news, too. Paul Kocher published the first timing attack against RSA in 1996 and the first power-analysis attack against DES in 1999. Within a decade, every smart card on Earth got re-engineered to mask its power consumption. The chips just leaked.

Post-quantum cryptography (the new generation of cryptography designed to survive quantum computers) inherits all of it.

ML-KEM, ML-DSA, Falcon, SPHINCS+, and HQC. Those are NIST’s post-quantum picks, meant to replace today’s encryption over the next decade. Every one has working side-channel attacks published against its reference code on GitHub.

KyberSlash (2023–2024) pulled private keys out of the most popular post-quantum libraries by clocking how long different inputs took to process. Falcon leaks its private key through tiny power-consumption variations while running. Dilithium has a step whose timing reveals the secret key. None touched the math. They broke the box.

Five channels do most of the work:

• Timing. Different inputs make the algorithm run at slightly different speeds. Measure carefully across enough samples, and the timing curve draws the key.

• Power. The chip drinks different amounts of current depending on what bits it’s computing. Scope on the power pin, average traces, the key reads itself off.

• Electromagnetic. Same as power, but the attacker doesn’t even need to touch the chip. An antenna nearby catches the trace. Air-gapped systems leak this way.

• Cache. Shared-cache systems leak which memory addresses a process touched. If a secret-dependent lookup lives in cache, another process on the same machine can watch.

• Fault. Inject a glitch (voltage drop, laser pulse, or clock pulse) at the right moment, the algorithm spits out a wrong output, and the difference between wrong and right often gives up the key.

A sixth one worth naming: the implementation choices upstream of the math. This could be bad random-number generators, key storage that survives a cold boot, or even memory that doesn’t get zeroed. None of these are the algorithm, and all of them break the system as completely as a math break would.

The procurement deck doesn’t usually mention this part. NIST tested algorithms… not so much the implementations. Side-channel resistance was a tiebreaker, not a gate, and the reference code on GitHub mostly isn’t side-channel-protected. Certified protected implementations exist, just in a much smaller universe than the code that compiles and runs.

The algorithm is the easy question. The implementation is where the work hides.

The math survives Shor’s algorithm (the famous quantum attack that breaks today’s RSA and ECC). Butttttt the power cord still snitches.

I — Investment Watch

• QuantWare closed $178 million Series B, the largest private round on record for a dedicated quantum-processor company. Intel Capital led, In-Q-Tel (the CIA’s investment arm) participated. CIA and Intel on the same European cap table is the part to notice.

• Quantum Motion closed $160 million Series C on a silicon-CMOS thesis: build quantum computers on the same boring chip process every classical chip uses. DCVC and Kembara co-led.

• eleQtron raised €57 million Series A led by Schwarz Digits, the tech arm of the German conglomerate behind Lidl and Kaufland. Three European mega-rounds in one fortnight (QuantWare, Quantum Motion, eleQtron) is the headline pattern.

• SkyWater shareholders approved IonQ’s $1.8 billion acquisition on May 8. A quantum company buying a U.S. trusted semiconductor foundry. First explicit quantum-acquires-foundry move in the U.S., and a CFIUS data point.

F — Field Notes

In any 2026 PQC migration conversation, the first question is which algorithm: Kyber or ML-KEM, Dilithium or Falcon? The procurement deck has a “PQC algorithm” column and a checkmark goes in it or it doesn’t.

That question’s wrong by one little layer.

A “systematization of knowledge” paper (basically: forty pages of senior researchers saying “here’s everything we know about this so far”) landed on arXiv this past week from Bertino et al. Their argument’s pretty clean. The algorithm’s the LAST question, not the first.

Before you pick an algorithm, you gotta pick an architecture. There are four flavors, not interchangeable:

• Symmetric-only. Keep AES and friends, deploy no new public-key crypto.

• Hybrid classical-plus-PQ. Classical and post-quantum running together; both have to break for the connection to fail. Most of the internet’s default migration path.

• Full PQ-PKI. Swap RSA and ECC out entirely for the NIST post-quantum picks.

• Information-theoretic. Cryptography whose security doesn’t depend on math being hard. Quantum key distribution is the textbook example.

Each architecture answers a different threat (harvest-now-decrypt-later, transit-only, decades-long confidentiality, regulatory mandate). Most enterprise environments are structurally legitimate for exactly 1/4.

In real audits the algorithm column gets filled in before the architecture question gets asked. The team picked ML-KEM because the vendor advertised it.

Nobody asked:

• if the threat overlay justified PQ-PKI?

• if a symmetric-only architecture would have done the job?

• if the implementation that shipped is the certified side-channel-resistant one or the reference build?

The order that actually works: architecture, then algorithm, then implementation, then certification. Each layer gets locked in well before the next gets touched.

When the migration plan in front of you has one column for “algorithm” and no columns for architecture, threat overlay, implementation profile, side-channel posture, or certification, what you’re looking at is the vendor’s plan with the client’s name on it (and that’s not the right incentive alignment).

T — The Tangent

Pull up any symmetric-key post-quantum migration plan and you’ll find one sentence carrying the whole thing: double the key length.

AES-128? Move to AES-256. SHA-256? Move to SHA-384. ChaCha20 was already 256-bit, sit down, you’re fine. Done. Symmetric side’s now closed with a shiny red ribbon. Most boards have signed off on exactly that paragraph.

The reasoning makes sense. Grover’s algorithm (Lov Grover’s 1996 quantum-computing trick) lets a quantum computer chew through symmetric keys in roughly the square root of the time a regular computer would. So your 128-bit key effectively feels like a 64-bit key to a quantum attacker. Double the bits, restore the margin, ship it with 25 years of consensus.

A paper out last week says the rule is… incomplete.

A four-institution team (Oxford, Germany’s BSI a.k.a. their NSA-lite for crypto, plus the German Aerospace Center and Bundeswehr University Munich) got something called Simon’s algorithm running on actual quantum hardware in April 2026. And, despite most assumptions, it wasn’t a simulation. IBM’s ibm_miami superconducting quantum chip, recovering symmetric keys from a real cipher with very few queries, reproducible across five experiments.

Quick context.

Simon’s and Grover’s are two different quantum algorithms. Grover’s chews through the key space the slow way (that’s the one your company’s AES roadmap is planning for). Simon’s looks for hidden repeating patterns inside the math the cipher uses.

When a cipher carries that kind of pattern in its structure, Simon’s runs WAY faster than Grover’s. Orders of magnitude faster, not “a bit” (pun intended). Grover’s halves the key length. Simon’s, on the right cipher, eats it whole for lunch.

The cipher the team attacked is Even-Mansour. Sounds obscure, until you realize it’s the structural kernel of one-round AES and the FX key extension. Real ciphers in real protocols sit right on top of it.

There’s a new map in town with three classes, instead of one.

• Grover-bounded. AES-256, ChaCha20, Serpent. Risk capped at Grover’s speedup. Doubling the key length works as advertised.

• Period-finding-vulnerable. Even-Mansour, one-round AES, FX, Chaskey, PRINCE, PRIDE, DESX, 2XOR, Elephant. These carry the algebraic structure Simon’s eats. Doubling the key length doesn’t reliably help. The cipher has to be redesigned or replaced.

• Structure-revealing modes. CBC, OFB, CFB, CTR, XTS. The mode matters as much as the cipher inside it; risk depends on whether the mode reduces to a period-finding problem.

The old “just double the key length” rule covers exactly one of those three classes.

The procurement question for every symmetric primitive in your stack just got bigger. You need the algebraic structure, which class it sits in, and the analysis behind it. If the vendor can’t answer, the roadmap has a column nobody filled in.

Sure, they recovered keys at N=3 and N=4; call that a toy if you want…

At N=5, the regular (non-quantum) computer they were using to set up the attack ran out of memory. On a server with 224 processor threads and nearly four terabytes of RAM. Surprisingly, the quantum chip wasn’t the bottleneck. Instead, it was the classical software prepping the attack. That constraint moves faster than qubit counts grow.

Same pattern as the H section. Asymmetric primitives leak through side channels the algorithm column doesn’t see. Symmetric primitives have algebraic-structure attacks the key-length column doesn’t see. The math survives Shor and the math survives Grover. The space around the math is where the migration plan has to be richer than one number per primitive.

See ya in two weeks!

Addie LaMarr

Founder, LaMarr Labs

⚛️

The Quantum S.H.I.F.T.

Issue #10 | April 28, 2026

S — Status Report (The Intelligence)

• A researcher just broke a 15-bit elliptic curve key on a publicly available quantum computer and won 1 Bitcoin for it. Project Eleven’s Q-Day Prize went to Giancarlo Lelli on April 24. He used a variant of Shor’s algorithm to derive a private key from its public key on real hardware. That is 512x larger than the last public demo seven months ago. Bitcoin uses 256-bit keys, so the network isn’t breaking this year. However, seven months ago the public record was 6 bits. Now it is 15, on hardware anyone can rent.

• IonQ published a buildable, end-to-end blueprint for a fault-tolerant quantum computer. The “Walking Cat” architecture maps a 110-logical-qubit machine to about 2,514 physical qubits in its dense configuration, not millions, and every hardware operation in the design has already been demonstrated on small trapped-ion devices.

• QuEra, Harvard, and MIT just hit a 2:1 physical-to-logical qubit ratio. For years the assumption was that one reliable qubit needed hundreds or thousands of unreliable ones holding it up. They got it down to roughly two. That is the kind of compression that quietly rewrites every roadmap built on the old number.

H — How It Works

What a 15-Bit Break Actually Means

Shor’s algorithm is the mathematical procedure that lets a sufficiently large quantum computer turn a public key into the matching private key, breaking the digital signature schemes that protect Bitcoin, Ethereum, and most of the internet.

Now let’s translate that to ~something we can understand~.

Imagine someone tells you they figured out the lock on a 4-digit suitcase. You shrug. There are only 10,000 combinations. A bored teenager could brute force that.

Then they tell you the lock on your front door has 256 digits.

You feel safer for about two seconds. Then they show you a video of themselves opening the suitcase by typing one combination, the right one, on the first try. Not guessing. Calculating.

That is what Shor’s algorithm does. It calculates the answer by exploiting the math under the lock to walk straight to it.

Lelli’s demo cracked a 15-digit version. Bitcoin uses 256. The gap sounds gigantic, and on classical hardware it would be. But the search space grows exponentially, not linearly. So the question worth tracking is “how fast is the floor moving up???”

In September 2025, the public record was 6 digits. In April 2026, the record is 15. That is 512x in seven months. On a publicly available quantum computer. Not a classified lab.

In the same two weeks Lelli ran his demo, IonQ published a blueprint that says a useful fault-tolerant quantum computer needs 2,514 physical qubits, and a separate team got the cost of one logical qubit down to roughly two physical ones.

Three things have to converge before a quantum computer breaks ECC at scale: enough physical qubits, an efficient enough algorithm, and cheap-enough error correction. Each one has its own field, its own funding curve, its own timeline. This month, every one of those numbers moved in the same direction.

That is what Google and Cloudflare meant when they pulled their full-PQC migration target to 2029, well ahead of the mid-2030s timelines almost everyone else is working from.

I — Investment Watch

• Sygaldry Technologies came out with $139 million across two rounds to build quantum-accelerated AI servers. Chad Rigetti’s new company. Breakthrough Energy Ventures led the Series A; Initialized led the seed. The bet: AI data centers are running out of power, and quantum hardware can speed up the math AI actually does.

• Quantum Art extended its Series A to $140 million. Bedford Ridge Capital led, with Hudson Bay, Poalim Equity, LIP Ventures, Wolverine Global, and IDA Ventures joining. The money goes into Perspective, their 1,000-qubit multi-core trapped-ion system. The phrase “global financial leaders” in the headline is the part to notice. Sovereign-bank and hedge-fund money are now in the cap table next to the venture firms.

• Quantinuum filed for a traditional IPO via Honeywell. Confidential S-1 submitted to the SEC on February 17, announced publicly April 22. The September 2025 capital raise priced Quantinuum at a $10 billion pre-money equity valuation. This is the first big quantum company to go public without a SPAC, and the SEC review is the first real public-markets stress test for the sector.

F — Field Notes

Every vendor pitch deck right now has a green checkmark next to “post-quantum cryptography.” Every press release has the phrase. Every renewal call includes some version of “we support PQC.”

Here is what most teams do not realize… Supported and enabled are COMPLETELY different states.

AWS, Cloudflare, Google, and Apple have all shipped PQC support. That part is real. What I keep finding is that the traffic still goes out classical. The vendor turned the feature on as an option.

The customer never opted in.

The board hears “quantum-safe” and the wire is still negotiating X25519 like it’s 2019.

The fix is one OpenSSL command, twenty-four hours of TLS telemetry, and a question to the vendor for the named-group preference order in production config (not the supported-list doc on the marketing page; those are two different files).

But almost nobody is running the check. Procurement’s reading the vendor’s claims. Security’s trusting procurement. The board’s trusting security. And the wire is going out classical because the dropdown was never set.

The HNDL exposure is hiding inside the gap between “we support it” and “it is on.” That gap’s where the next two years of quietly harvested traffic lives.

If you are a CISO and you have not personally watched a TLS handshake on your own outbound traffic this quarter, you do not actually know what your posture is. You know what your vendor told you it is.

T — The Tangent

Almost every PQC roadmap is built on this assumption:

Public cryptanalysis tracks classified cryptanalysis closely. If a serious break existed, we would have a rumor. We’d see a leak. The math is too pretty to keep secret for long.

In the early 1970s, three GCHQ researchers built what would later be called public-key cryptography. James Ellis proposed the concept in 1970. Clifford Cocks worked out the RSA equivalent in 1973. Malcolm Williamson worked out the Diffie-Hellman equivalent in 1974. All of it was classified.

The public versions came later. Diffie and Hellman published key exchange in 1976. Rivest, Shamir, and Adleman published RSA in 1978. GCHQ kept their version classified for another twenty years on top of that. Ellis died on November 25, 1997. Cocks gave the public lecture revealing the prior GCHQ work on December 18, 1997. That was a gap of twenty-three days. Ellis spent his entire career unable to tell his colleagues he had invented the math their work would depend on.

Twenty-four years of asymmetry. Public researchers thought they were at the frontier. They were several years behind, and the gap felt like nothing because nobody knew it existed.

Now apply that to right now.

A researcher just broke a 15-bit elliptic curve key on a publicly available quantum computer. Public cryptanalysis is showing its work in the open, on commercial hardware, with a Bitcoin bounty as the receipt. NSA, GCHQ, MSS, and Unit 8200 are running it on whatever they have built and whatever they have not declassified.

The honest question for any CISO sitting on a 2030 or 2035 migration timeline is the GCHQ question rephrased. If a state actor had a working attack on ML-KEM or ML-DSA today, would you expect to know about it? When? From whom?

“Harvest now, decrypt later” is the modern version of the GCHQ vault. The traffic you put on the wire today is the traffic that gets read when the asymmetry collapses. The collapse will not announce itself. The 1997 declassification did not.

The roadmap that assumes you have years because you have not seen a public break is making the same mistake the public-key researchers made in 1976.

— Addie LaMarr

Founder, LaMarr Labs · Post-quantum cryptography advisory

Request an initial briefing →

Issue #09 | April 15, 2026

S — Status Report (The Intelligence)

• AI just found a faster way to break RSA encryption, and Cloudflare immediately moved their deadline up by three years. A startup called Oratomic published research showing that with AI-assisted algorithm design, quantum computers need way less power to crack RSA than anyone had planned for.

• Scientists can now measure quantum data loss 100x faster than before. Quantum computers leak information constantly (decoherence), and diagnosing how fast just got dramatically easier… which means fixes can happen dramatically faster too.

• The global quantum market is projected to hit $3 billion by 2028. QED-C’s 2026 industry report says the market will double in the next two years, with defense and enterprise security doing most of the lifting.

• China’s quantum funding is surging back after a quiet stretch. The US-China quantum race is the actual subtext behind every PQC policy move happening in Washington right now.

H — How It Works

AI-accelerated algorithm discovery is the process by which machine learning models identify previously unknown mathematical shortcuts within established cryptographic systems.

Now let’s make that ~achhhhtually understandable~.

You know that one kid in class who somehow already had all the answers before the exam was even handed out?

Here’s the secret: bro didn’t study harder or understand the material better. They found the answer key.

That is literally what AI just did to RSA encryption.

Oratomic pointed an AI at the RSA problem and said: find the shortest path to the answer. The AI didn’t gaf about the decades of research humans had carefully stacked up. It scanned the entire solution space, found a mathematical shortcut nobody had spotted, and handed it over.

Resulting in a 20-page paper that caused Cloudflare to move their internal deadline up by three years.

So what’s the catch?

Every “we have until 2032 to prepare” timeline was built on one assumption: that finding new ways to break encryption moves at human speed. Stuff like PhD teams, peer review, conferences, and years per breakthrough. Yadda, yadda yadda.

Buuuuuut AI has never moved at human speed.

We have no idea how many answer keys like Oratomic’s are sitting in a closed lab right now, waiting to be published… or secretly exploited.

I — Investment Watch

• Q-Factor came out of stealth with $24 million to scale neutral atom quantum systems. Intel Capital, NFX, and TPY Capital are in — neutral atoms use laser beams instead of extreme cooling, which could make quantum dramatically cheaper to run if it works.

• Peak Quantum raised €2.2 million pre-seed to build a European quantum hardware manufacturing line. Europe does not want to depend on US or Chinese hardware for its critical infrastructure, and this is that effort in its earliest form.

• IonQ and the University of Maryland expanded their National Quantum Laboratory with $7.5 million focused on quantum networking. Forget computing. Networking. The pipes are getting funded… which means someone is actually planning for a world where quantum systems have to talk to each other.

F — Field Notes

The conversation I keep having goes like this.

Security team: “We know quantum is coming. We just can’t get leadership to prioritize it.”

Leadership: “We’ll deal with it when it’s a real problem.”

And then Cloudflare does this…

A company most executives in the room actually recognize, quietly announcing they were wrong about their own timeline and changing their plan.

That’s the language that actually moves a budget… A name they know, making a decision they can’t ignore.

Use it if you’re trying to convince your leadership to actually make some moves.

T — The Tangent

I don’t give a HOOT how many times people ask me “how many stable qubits exist rn?!?!?!” like its some kinda ‘gotcha!’ that proves quantum should be ignored.

When I talk about Q-Day (or the Quantum Threshold… because Q-Day is a myth), people always drop the “qubit count” like it’s a conversational kill-switch. They think citing a hardware roadmap is a “gotcha” that proves we have decades of safety. I hate to say it but that’s cope. It’s the digital equivalent of saying a fort is safe because you haven’t seen the enemy building a ladder… meanwhile, the enemy just found a way to walk through the walls.

They keep imagining it’s a hardware race while it’s actually a race of suppression.

The Qubit Smoke Screen

We’ve been conditioned to treat quantum progress like a public construction project. We wait for the press release, “QuAnTuM sUpReMaCy” headline, or the benchmark. But in the world of high-stakes intelligence, stability is an unlikely variable, never a guaranteed prerequisite. If a nation-state finds a mathematical shortcut that allows “noisy,” unstable hardware to crack a specific key, they aren’t going to announce that. It would defeat the whole purpose. They’re going to use it.

Strategic Silence is the Real Weapon

In this dumpster fire of a political climate, the most dangerous advancement is the one that stays quiet.

• State Suppression: If a breakthrough is made in a government-funded lab, it gets classified. Like Hedy Lamarr’s frequency hopping tech that never saw the light of day until after she passed.

• The Exploit over the Nobel: There is zero incentive for a global power to “win” the quantum race publicly during these unprecedented times. The prize = the ability to read every encrypted message sent in the last decade. You won’t brag about having the skeleton key unless you want everyone to change the locks… you simply open the doors to loot what’s inside.

We’re Living in the Shadow

The “we have time” crowd is looking at the scoreboard. They don’t realize the game is being played in a backroom they aren’t allowed to enter. Forget the idea of a “Quantum Dawn” aka a bright, public moment where everything changes. We’re likely already living in the Quantum Shadow. The moment the encryption was broken, the smartest people in the room stopped blabbing.

Cloudflare read one public paper and moved three years. Imagine how far the people who suppress the papers have already moved.

I just touched down from a trip to Belgium, where I was delivering a keynote at CyberNova on why the "Q-Day" narrative is a myth. The timing’s super eerie… I was literally mid-flight when this Google news hit my inbox, proving that while the "apocalypse" is hype, the math is moving way faster than anyone expected.

So, Google’s Quantum AI team (led by some very smart people with very impressive resumes, Dr. Ryan Babbush and Dr. Hartmut Neven) just dropped a paper.

Usually, academic papers are boring. They have charts that look like spaghetti and words like “superposition” that people use at cocktail parties to sound smart.

But this one’s built DiFfErEnT. It basically says: “Hey, ya know that ‘Quantum Apocalypse’ everyone said was 10+ years away? Yahhhhhhh... it’s actually right around the corner. And we found a shortcut.”

Google revealed they can now break 256-bit Elliptic Curve Cryptography (the stuff that keeps your Bitcoin safe and your DMs “private”) using 20x fewer resources than we previously thought.

It’s like thinking you needed a 1,000-man army to rob a vault, and then Google realizes you actually just need 50 guys and a very specific type of grenade.

Previously, the smart people thought we needed millions of qubits to break Bitcoin.

Google’s new blueprint says:

• Logical Qubits needed: Less than 1,200.

• Physical Qubits needed: About 500,000.

• Time to crack: Roughly 9 minutes.

Why 9 minutes? Because Bitcoin takes 10 minutes to confirm a block. If a hacker can derive your private key in 9 minutes, they can front-run you & steal your money before the network everrrrr knows what happened.

The Secret Blueprint

But it gets weirder: Google didn’t publish the actual circuits.

Usually you show your work and give everyone the recipe. But Google looked at this and said: “This is way too dangerous to give to the internet.” Instead, they used something called a Zero-Knowledge Proof. It’s a way of saying, “I have the answer, I can prove I have the answer, but I’m not showing you how I got it because you’ll probably use it to do something stupid (like collapse the global economy).”

It’s the first time we’ve seen “Responsible Disclosure” for math. They’re treating a physics discovery like a government secret.

Why Should You Care? ($$$$$$)

If you have Bitcoin, Ethereum, or a bank account, your stuff is locked with ECC-256.

• 7 Million Bitcoin are currently sitting in “exposed” addresses. That’s 30% of all BTC ever made. To a quantum computer, that’s just free money sitting on a sidewalk waiting to be taken.

• Governments are panicking because their VPNs and secure comms use that same math.

TL;DR:

1. Q-Day’s closer: The “2029” timeline just got a lot more aggressive than the 2032-2035 estimate.

2. The Blueprint’s Hidden: Google’s gatekeeping the “weapon” under lock and key... for now.

3. Migration’s Non-Negotiable: If your company isn’t looking at Post-Quantum Cryptography right now, you’re gonna be utterly blindsided by this threshold.

Google didn’t build the machine yet, but they finished the map for it. And the destination is a looooooot closer than we thought.

Stay tuned for the next Quantum SHIFT,

— Addie

The Quantum S.H.I.F.T.

Issue #06 | March 3, 2026

Status Report (The Intel)

• Google launches Merkle Tree certificates to keep the post-quantum web fast

  Instead of sending massive, slow quantum keys every time you visit a site, Chrome’s testing a system where one Tree Head verifies millions of certs at once. This means you get quantum security without 2005 era dial up speeds.

• Pasqal and Welinq launch a project to build the Quantum Ethernet

  They are using light to connect neutral atom processors, effectively building a network for quantum chips. This is how we move from a single, finicky computer to a scalable data center where chips actually talk to each other.

• NIST and QED-C finalize a plan to shrink quantum spaghetti wiring

  Researchers delivered a blueprint to consolidate the massive mess ‘o cables and electronics surrounding quantum computers into standard data center racks. This moves the tech out of science experiment territory and into something that looks like ~actual~ infrastructure.

• Tectonic Labs releases a browser wallet with post-quantum signatures

  The PQ Wallet uses Falcon-512 math to protect Ethereum based assets from being drained by future quantum attacks. It’s one of the first usable tools for people who don’t want their crypto keys harvested now and cracked later.

How It Works

Quantum Sensing

Imagine you’re standing in the middle of a packed football stadium. Somewhere in the very last row of the nosebleed seats, a guy opens a tiny jar of peanut butter.

A classical sensor’s like a human nose. You aren’t smelling anythingggg. You’re lucky (or maybe not) if you can smell the guy sitting next to you. It’s just too dull to pick up a tiny change in the air from that far away.

A quantum sensor is like a bloodhound on steroids. It both “smells” the peanut butter and it picks up a single, solitary molecule of it from across the field and knows exactly which seat it came from.

In the real world, we use atoms instead of dogs. Because these atoms are kept in a “quantum” state, they are ridiculously sensitive to the world around them. If a magnetic field shifts by a teeny fraction of a hair, or if the gravity changes because there is a hollow pipe buried ten feet underground, the atoms “twitch.”

We are using that twitch to see through walls, find oil without digging a single hole, and navigate submarines without needing a GPS signal from space.

This concept is the epitome of “Work smarter, not harder.” You’re simply using a much more sensitive “nerve ending” to feel the world.

Investment Watch

• Finland’s IQM announces $1.8B SPAC merger to list on the Nasdaq

  Europe’s top quantum hardware firm is jumping into the US public markets to fuel their build out of full stack superconducting chips. This is the first time a major European quantum player is hitting the big leagues to challenge the US giants.

• D-Wave revenue surges 179% as industrial bookings hit record highs

  D-Wave reported a massive jump in sales and $30 million in new bookings this month alone. They are successfully selling “optimization” (solving logistics nightmares for big companies) today while they build their next gen hardware on the side.

Field Notes

Most people think “quantum ready” means picking the right math. It’s not.

The real bottleneck’s the hardware hangover. PQC keys are fatties… we’re talking 10x to 100x the size of the RSA keys we’ve used for decades.

If you’re running a high speed network or thousands of tiny IoT devices, your infrastructure might literally choke on the “secure” packets. You’re trading security for massive latency, and for some industries, that lag’s a death sentence.

But everyone is freaking out about the “math” when the fix is actually Crypto Agility.

Most people don’t realize it’s that simple. It just means building your system like a Lego set instead of a solid block of concrete.

If ya hardcode your encryption into your app, you’re stuck with a “concrete” problem when the wind changes.

If you use a modular layer, you can swap a weak algorithm for a strong one in an afternoon without rewriting a single line of core code.

Agility makes this process a zillion times less of a headache.

The Tangent

For the last decade, every tech bro with a podcast has been trying to convince you that we’re living in a simulation.

They look at the “glitches” in quantum physics and swear we are just characters in a massive game of The Sims running on an alien supercomputer.

New mathematical proofs from February 18th just stuck a needle in that balloon. Researchers have shown that the fundamental laws of reality are non-algorithmic.

This is a fancy way of saying that reality does not follow a step by step recipe.

Most people assume the universe is like a player piano… it can only play the notes that are prepunched into the paper roll. If it isn’t in the code, it doesn’t happen.

But these new proofs show that the quantum world is actually playing live jazz.

Because quantum systems can exist in “superposed” logical states, they make “jumps” to truths that literally cannot be calculated or proven by any sequence of code.

This is why “Artificial Intuition” is the next frontier for quantum computing: we’re moving past building better calculators and trying to build machines that can “feel” the next note.

This means the universe is an original.

Even if we built the ultimate quantum computer, it could never fully simulate a piece of fruit, let alone your consciousness, because the “source code” of reality isn’t code at all. It’s something deeper that exists outside of logic and math.

So, stop worrying about being a line of script in a hard drive.

You’re much more complicated than an algorithm. It’s a relief, but also a little terrifying when you realize there’s no quit button.

See you in two weeks!

Addie LaMarr

Note: I’m building Part 2 of my CyberNova 2026 keynote, so I wrote a 0-to-100 Quantum Threshold explainer I can pull from. It’s more in-depth than what I normally post, because the point is to give decision makers the whole scope of the quantum transition and why most timelines are going to lie to you.

The Day Digital Proof Stops Working

Quantum changes what “proof” even means and you may not get a warning.

Most people hear “quantum computing” and think, “Cool. Faster computers.”

But that’s not the real danger.

Quantum computing changes the rules entirely.
It makes some of the math we rely on for digital trust stop working the way we assumed it would.

Here is the definition to remember:

The Quantum Threshold is the moment the math behind modern digital security stops being reliable.

Some people call it Q-Day.
That’s shorthand for the day a quantum computer becomes powerful enough to break today’s encryption and fake today’s digital proof.

You probably won’t get a clean announcement, but you’ll certainly see the anomalies.

The Corn Maze Problem

Alright, I want you to stop picturing a computer chip for a second.

Instead, imagine a corn maze.

I mean a serious one that has towering walls. You can’t see over them or the exit.
And every turn looks the same, so the second you commit to a path, you’re basically gambling.

Now imagine you’re the one inside it.

You start running.

You take a left. You sprint.
Dead end.

So you turn around and run back.
You go right. You sprint again.
Another dead end.

You keep doing this over and over. Until your legs burn and you start getting annoyed.
You start thinking, “There has to be a smarter way.”
But there isn’t. Not for you.

That’s the whole point of the maze.

It’s designed to make finding the exit brutally hard if all you can do is try paths one at a time.

That’s what modern cryptography is: a maze made of math.

And the “exit” is the secret key or the private key. The one thing nobody is supposed to be able to find.

Right now, our entire digital world runs on one belief:

The maze is too hard to run.

A classical computer is basically you in that maze, except it has unlimited stamina and it can run insanely fast.

So yes, it can try billions of paths per second.

But it’s still doing the same basic thing you’re doing.

One path at a time.

Now watch what happens when a quantum computer enters the maze.

Instead of running, it turns into a thick, silver mist.

I want you to see it like fog pouring in under a door, spreading everywhere at once, filling corridors, dead ends, hidden turns, all at the same time.

The mist is everywhere. And while it’s everywhere, it’s testing huge numbers of paths in parallel, and then it uses a physics trick to clean the maze up.

The wrong paths cancel out.
The right path gets thicker.

So when you finally ask the quantum computer, “Okay, where’s the exit?” it’s not guessing the same way you were.

It’s pointing to the one path that survived.

Quantum changes the rules of the maze by making the maze irrelevant.

And that is the Quantum Threshold.

The Four Pillars of Reality

Most quantum conversations get stuck on this idea:

“Private data might be exposed.”

Yes. That’s bad.

But it’s not the scariest part.

This is:

Proof becomes disputable and forgeable.

Because once proof is forgeable, can your business still prove what’s real?

Now, let me translate the technical stuff into business realities.

Quantum threatens the mechanisms that let systems prove reality:

Digital signatures.
This is the thing that says, “This email is really from our CEO, this software update is really from Microsoft, and this contract was actually signed by our vendor.”

If that breaks, you get believable forgeries at scale.

Authentication.
This is the thing that says, “This is actually your customer logging in, employee approving payroll, and your finance lead approving a wire transfer.”

If that breaks, access stops being a lock.

Integrity.
This part says, “This record wasn’t altered, audit trail is accurate, and transaction history is the truth.”

If that breaks, you can’t confidently answer basic questions like:
“Who approved this?”
“When did it change?”
“What did it say yesterday?”

Privacy.
Yes, this is the encrypted data part. Your customer data, M&A docs, pricing, legal strategy, and product roadmap.

But even if privacy’s the only thing you think you care about, the moment authenticity and integrity wobble, you’re losing credibility.

These are the invisible foundation of business.

So when proof breaks, it looks like this:

Your finance team gets an invoice that is perfectly signed and formatted but 100% fake.
Your bank accepts it because the “proof” checks out.
Your money leaves.
Your vendor swears they never sent it.
And you can’t prove they did.

That is what I mean by “reality gets slippery.”

But this is what matters for strategy: If someone can forge proof, the advantage is so extreme that they don’t announce it.

And we have to look to history for the clearest example of that logic.

The Lesson From Alan Turing

During World War II, the Germans were sending encrypted radio messages every day. They sent orders, coordinates, and timing. The kind of messages that decide which ships make it home.

Britain’s best chance was a place most people had never heard of: Bletchley Park. Inside, Alan Turing and a small team of codebreakers were racing the clock to break Enigma, because every hour they couldn’t read those messages meant more people dying in the dark.

When they finally crack it, they can’t celebrate. It’s a new kind of pressure. Because now they can see what’s about to happen before it happens.

This next part is from the movie version (The Imitation Game). It’s dramatized, but it captures the strategic truth perfectly.

In the film, someone on the team reads a decrypted message and realizes his brother’s on one of the ships marked for attack. He goes cold, wants to run to a phone and shout. He desperately wants to save him.

But the leaders stop him.

Because they understand the real game.

If they act too obviously, the Germans will notice the pattern and realize Enigma is broken. They’ll change the whole system and build a new maze.

So in the movie, the team does the unthinkable.

They stay silent, and let the attack happen, all while protecting the secret.

Brutal. And completely rational.

Because once you can break cryptography, the advantage is so extreme that the first rule is simple:

You don’t tell anyone you have it.

The Quietest Breakthrough Wins

Now translate that into today.

In the U.S., we invest in quantum and we talk about it loudly. The 2018 National Quantum Initiative put $1.2 billion behind quantum research, and a lot of the ecosystem is public by design because public wins impress shareholders.

China plays a completely different game. Multiple credible analyses put China’s public quantum investment around $15 billion and describe a more state directed, secretive approach to development and disclosure.

So ask yourself this.

If a nation state crosses the Quantum Threshold, do you really think they throw a press conference?

If you can quietly forge proof, quietly decrypt secrets, quietly impersonate trusted systems, the smartest move is to keep your mouth shut.

So don’t plan for a clean public Q-Day. Especially not in this global climate.

The first sign may be downstream failures and silent advantage:

• fraud that keeps passing verification

• deals that fall apart for reasons nobody can prove

• “impossible” incidents that keep happening

• competitors moving like they already know your next move

And because only nation states have the resources to make this real, I started paying attention earlier than most… starting with a night in Japan in 2016.

The Night I Stopped Believing in Linear Timelines

In 2016, I was stationed in Japan. Once a month I went to an event with the Okinawa Institute of Science and Technology (OIST) called Nerd Nite.

It was my favorite night of the month.

PhDs and researchers would go on stage with a beer, nerding out over their research obsessions like zombie ants.

One night a researcher gives a talk on quantum computing. Someone asks the question everyone asks:

How far away are we?

He says, “Maybe 50 years.”

And the room breathes a sigh of relief.

Fifty years means it’s not my generation’s problem. I went home thinking quantum was a story for another time.

Then ten years later, that 50-year window doesn’t feel like 40 years left.

It feels like the whole timeline collapsed.

Because now, a lot of experts are putting “cryptographically relevant” quantum machines inside a 10-year planning window, with many estimates landing in the five to 10 year range.

In fact, the EU has a coordinated roadmap that effectively makes 2026 the starting gun for rolling out post-quantum cryptography in priority systems, and U.S. federal guidance has agencies inventorying quantum-vulnerable cryptography and beginning the transition planning process now.

Breakthroughs stack, compound, and jump.

And leaders get punished for planning on the old clock.

The Proof Wobble Era

So here is what I want you to take with you:

The Quantum Threshold won’t feel like a moment where everyone agrees it’s here.

It’ll feel like a period where the world starts glitching out.

Where proof starts wobbling, and reality becomes easier to dispute.

And the people who moved early look paranoid, until suddenly they look inevitable.

That is why if you’re waiting for certainty, you’re waiting to be late.

The Quantum S.H.I.F.T. Issue #05 | February 17, 2026

Status Report (The Intel)

• Spanish team (CSIC) claims Majorana breakthrough amidst skepticism Researchers from CSIC and QuTech published a paper in Nature demonstrating a new "quantum capacitance" technique to read Majorana qubits, though critics remain vocal that this small-scale lab success doesn't prove that Microsoft’s massive, commercial "Majorana 1" chip is actually functional or free of "fake" signals.

• White House drafts “whole-of-government” Quantum Executive Order

  A new draft directive titled “Ushering in the Next Frontier of Quantum Innovation” has surfaced, aimed at centralizing federal quantum R&D and building a dedicated quantum computer for scientific discovery.

• Google issues urgent “Borrowed Time” warning for current encryption

  Google’s leadership is pushing for immediate migration to post-quantum standards, warning that “harvest now, decrypt later” attacks are already happening and the window to act is closing fast.

• NSF launches $100M quantum infrastructure plan

  The National Science Foundation is investing up to $100 million to establish 16 open-access sites across the U.S. to accelerate quantum fabrication and workforce training.

• Leiden University unveils “Tortilla” quantum microscope

  This new multi-modal microscope can image temperature and magnetism at the nanoscale without disturbing delicate quantum states, which helps speed up the manufacturing of quantum chips.

How It Works

The Maze Runner: Classical vs. Quantum

Imagine you’re looking for a $1,000,000 prize hidden inside a massive corn maze.

A classical computer is like you sprinting through the maze. You hit a dead end, sigh, turn around, and try a different left turn. It works eventually, but it takes a ton of cardio and way too much time.

A quantum computer is like a thick mist that rolls into that same maze. Instead of picking one path, the mist just expands. It flows into every single path simultaneously. It isn’t “trying” the paths one by one; it just exists in all of them at once. The second that mist touches the exit, the whole cloud “collapses” onto the right answer.

You aren’t working harder. You’re just being everywhere at the same time.

Investment Watch

• IonQ raises $2 billion more in equity

  IonQ reported a massive cash position of $3.5 billion after a successful $2 billion raise, giving them the runway to scale their platforms despite the current heavy R&D costs.

• Quantum revenues projected to hit $2 billion this year

  Market analysts expect the quantum sector to reach the $2 billion mark in 2026, primarily driven by spending in the defense and aerospace sectors.

Field Notes

The biggest mistake I see is treating the quantum transition like a rando “tech project” that belongs in a basement with one lonely sysadmin. It is NOT.

This is a business continuity crisis.

If your leadership team thinks PQC is just a lowly software patch, you’ve already lost.

You need the board’s attention because this transition requires actual capital, and they are usually the primary bottleneck.

The only way to move them is to stop talking about “qubits” and start talking about “cost per month.”

Show them the price tag of doing nothing… the liability of harvested data and the inevitable rush fees of a forced migration, and watch how fast the “tech problem” becomes a strategic priority.

The Tangent

The "dark side" of quantum is hidden in the millions of pages of the Epstein files. I have a full length deep dive coming, but in the meantime… here’s the gist:

When you look at the names in those wretched docs between top physicists, tech titans, and behavioral scientists, you see a pattern.

They were OBSESSED with data. Why?

Because whoever controls the data right now is building a "Digital Twin" of you.

Every app, email, and "deleted" message is being harvested to create a perfect replica of your psyche.

And they’re also collecting your encrypted data (think: cloud syncs) to decrypt later.

I like to call it Quantum Necromancy: the ability to use future computing power to "resurrect" your digital ghost from the traces you leave behind today.

Others simply call it harvest now, decrypt later… but I don’t think it conveys the reality of whats happening or the specificity of it being applied to digital twins.

Let’s put it into perspective…

Right now, Meta uses "ad sequencing" to prime you.

They might show you a "cute" reel of a puppy, followed by a "horrifying" news clip, just to put your brain in the exact state of vulnerability needed to make you buy a pair of shoes you don't want.

Quantum Necromancy takes this to a terminal level.

By cracking your archived, encrypted data, these machines become the engine that runs ridiculously fast simulations on your digital ghost.

They use that corn maze to find the exact sequence of events that forces you (and society at large) to change your mind or take an action you wouldn't otherwise.

Significantly faster than the current tech titans curate your algorithm.

It’s the difference between an ad and an automated puppet string.

Privacy today is so much more than "hiding secrets."

It’s really just preventing someone from owning the literal blueprint to your autonomy.

If you aren't reasonably encrypting data now, you’re just handing them the keys to controlling your future self.

See you in two weeks!

Addie LaMarr

The Quantum S.H.I.F.T.

Issue #04 | February 3, 2026

Status Report (The Intel)

• Davos leaders warn that waiting for quantum is no longer an option

  The World Economic Forum just wrapped up with a blunt message: quantum is becoming to strategic infrastructure, and if you aren’t planning for the “Y2K moment” of encryption now, you’re already behind.

• Stanford researchers build a tiny “light trap” for qubits

  They created miniature optical cavities that can read hundreds of individual atoms at once, which is basically the high-speed internet upgrade quantum computers need to finally scale up to millions of qubits.

• Palo Alto Networks launches “Quantum-Safe” security solution

  One of the biggest names in firewalls just started selling a way to automate the swap to post-quantum math, proving that “harvest now, decrypt later” is officially a boardroom-level problem.

• D-Wave shifts to Florida and shows off 10x faster missile defense

  The company is moving its HQ to Boca Raton and just proved their hybrid tech can plan missile intercepts ten times faster than normal computers, making this a massive win for national security tech.

• WISeKey plans the first post-quantum secure robot launch

  They are connecting a robot to a satellite constellation using quantum-resistant math to ensure that future “smart” machines can’t be hijacked by a quantum computer.

How It Works

The Supermarket Checkout (Quantum Parallelism)

Imagine you’re at a massive grocery store.

It’s the day before Thanksgiving. (yikes on bikes)

There are 50 checkout lines, and they’re alllllllll packed.

In the world of “classical” computers (aka the laptop you’re using right now), you are just one person.

You have to pick one line, stand in it, and wait.

If you pick the slow line, you’re stuck.

You can only be in one spot at one time.

A quantum computer is more like a ghost.

Instead of picking one line, you exist in every single line at the same exact time.

You don’t have to guess which line is the fastest because you are exploring all 50 possibilities simultaneously.

The moment one cashier hits “Total,” the versions of you in the slow lines just vanish into the void.

You instantly finish your shopping at the fastest possible spot.

Quantum computers go “wider” by trying every answer at once.

Investment Watch

• EU launches €50M SUPREME consortium

  The European Union put up roughly $59 million to industrialize superconducting quantum tech. They are tired of watching the US and China lead the pack and want to build their own “sovereign” quantum hardware.

• Pentagon boosts cyber budget to $15.1 billion

  The 2026 US defense budget includes a massive jump in funding specifically aimed at “quantum-resilient” systems. This is a huge signal to contractors that if your tech isn’t quantum-safe, the government isn’t buying it.

• Nu Quantum secures €9.75M for “Entanglement Fabric”

  The Spanish government is backing this UK-based startup to build a networking hub in Madrid. They are trying to create a way to link separate quantum computers together so they can work as one giant super-brain.

• IonQ’s $1.8 billion SkyWater acquisition stirs the market

  While quantum stocks have been a bit of a rollercoaster lately, IonQ’s massive move to acquire SkyWater is seen as a play to bring their manufacturing in-house and lock down more government defense contracts.

Field Notes

Pls stop panicking…

I keep seeing people freeze up because they think “Quantum” means they have to rip out every server and replace them with liquid-nitrogen-cooled boxes.

I’m even getting frantic questions about “Quantum VPNs.”

(I just recorded a video on this that drops in a few weeks: but the short answer is you probably don’t need one).

Most people are getting scared of the buzzword.

They think they’ll be forced into incredibly expensive hardware setups or “Quantum Key Distribution” just to stay safe.

Here is the truth: For 99% of you, this is a software update.

Think math problem, not construction project.

Most of the time, you can keep your normie hardware and just switch out the crypto for the new standards.

It’s much more like updating your browser, and a lot less buying a new laptop.

Focus on taking the first baby step (your data inventory) and let the researchers worry about the exotic hardware.

The Tangent

We used to draw a hard line:

Computers are digital.

Biology is analog.

We thought DNA was just a boring blueprint for things like eye color or whether you’ll be 6’ tall.

Then the Epstein files dropped.

Nearly 3 million pages of “what the helly” were released this week.

Buried in there is some truly bizarre dialogue about “biological encryption.”

source

The documents mention theories that every human has innate “encryption keys” and “decryption keys” encoded into their very being.

To be honest, it sounds like total sci fi nonsense from a guy way too obsessed with transhumanism.

But the unsettling part is that science is actually proving it out. 🫠

This week, researchers at ASU showed how “DNA Origami” can be used to create biological encryption keys.

It turns out your DNA is a blueprint AND a high density hard drive.

That strange talk in the files was a dark preview of a future where your own biology is the ultimate post-quantum firewall.

Talk about a security update you cannot uninstall.

The Quantum S.H.I.F.T. Issue #03 | January 20, 2026

Status Report (The Intel)

• Rigetti locks in an $8.4 million deal with India’s top tech agency. India is officially buying into the on-prem hardware game instead of just renting cloud time from the big guys.

• IBM doubles down on the 2026 Nighthawk processor. Big Blue claims this is the year quantum computers actually beat standard supercomputers at solving real world probs instead of just lab experiments.

• G7 financial groups release a coordinated quantum roadmap. The world’s biggest economies are finally admitting that our current banking security is a house of cards once the big quantum chips arrive.

• US Senate introduces the National Quantum Initiative Reauthorization Act. Congress actually moved fast (on Jan 8th) to dump billions into making sure we don’t lose the quantum race to China (who has already outspent the US by ~10x)

How It Works: Quantum Error Correction

In stuffy academia terms, Quantum Error Correction (QEC) is a set of algorithms and hardware techniques used to protect quantum information from errors caused by decoherence and other “noise” from the environment.

Because individual qubits are incredibly fragile, QEC maps one “logical” qubit of data onto many “physical” qubits to ensure the calculation stays on track.

Now, let’s make that ~achhhhtually understandable~.

Think about sending a risky text after 2 AM + 4 tequila shots.

Your thumbs are hitting three letters at once. Your brain’s lagging.

If you send that text as is, your “just a friend” gets a string of gibberish that looks like a cat walked across your keyboard.

Quantum Error Correction is like having a group chat of 50 sober friends who proofread your text before it actually hits “send.”

Because qubits are “drunk” by nature (they flip or lose their data if a flea sneezes in the next room), we can’t trust a single one to hold a thought.

Instead, we bundle a bunch of messy physical qubits together into one “Logical Qubit.”

These 50 friends look at the messy input.

If 49 of them see something that looks like “pizza” and one outlier thinks it says “ex,” the group ignores the drunk guy and fixes the message to say “pizza.”

So what’s the catch?

This is exactly why stable quantum computers are taking soooooooooooooo long to build.

To get one single “Logical Qubit” that actually works, you might need 1,000 or more shaky physical ones to babysit it.

It’s kinda like needing an entire stadium of people just to send one coherent tweet.

The overhead is literally insane, but it’s the only way to keep the quantum magic from becoming a dumpster fire the second it touches the real world.

Investment Watch

• Quantinuum files for a confidential IPO. Honeywell is spinning off its quantum darling. This is a massive signal that the “Quantum Winter” is thawing and the public markets are ready for real hardware companies.

• Project Eleven raises $20 million Series A. Led by Castle Island Ventures, this funding focuses on fixing the security of digital assets. They are essentially building the “bulletproof vest” for blockchains before the quantum “bullets” are even manufactured.

• Equal1 secures $60 million to use old-school tech They are building quantum computers using the same semiconductor tech we use for phone chips. It matters because it means we can mass produce these things in existing factories instead of building custom labs from scratch.

Field Notes

Most companies are currently suffering from a massive case of “The AI Distraction.”

A huge PwC study released this morning at Davos confirmed the nightmare.

A full 56% of CEOs admit they have seen zero financial return from their AI investments so far.

Imagine the billions of dollars being set on fire right now.

Executives are dumping cash into LLMs because of FOMO.

They’re embedding chatbots where customers don’t want them and employees don’t need them.

It’s the ultimate shiny object syndrome.

But while they’re chasing the FOMO, they’re also ignoring a fundamental truth: Being “AI-advanced” does not mean you are going to survive.

In fact, every terabyte of sensitive data you feed into these expensive AI models today just creates a bigger target for “harvest now, decrypt later” attacks.

You’re effectively building a gold plated vault but leaving the door WIDEEEE open for anyone with a future quantum key.

For years, cybersecurity has been treated like a dorky “cost center.”

It was the tax you paid to stay in business.

But the quantum transition changes the math.

This is the first time in decades where we have a clear mandate to rip out and replace the entire basement of the internet.

For the founders and leaders listening: this is your differentiator.

The companies that win the next decade will be the ones that provide the “unbreakable” foundation.

We are moving into an era where Security = Credibility.

If you can prove your infrastructure is post-quantum secure, you become the only trusted player in a room full of people who are still using screen doors to protect their digital gold.

The Tangent

Is it true that world’s overlords leaders are losing sleep over quantum?

Abso-freaking-lutely.

If you look at Davos this week, the vibe has shifted from “Quantum is a cool science project” to “Quantum is a national security emergency.”

The World Economic Forum and major tech players have officially branded 2026 as The Year of Quantum Security.

While the general public is still arguing with ChatGPT, the lovely people who actually run the world are holding closed-door sessions on how to keep their communications from becoming an open book.

It’s a little unsettling to watch the most powerful people on Earth suddenly get very nervous about their own digital vaults.

They’re finally waking up to the reality that a border is useless when the key to your EnTiRe national infrastructure is being calculated in a lab thousands of miles away.

When the WEF stops talking about the “metaverse” and starts talking about “Harvest Now, Decrypt Later” risks, you know the party is officially over for legacy encryption.

That’s it for this week, folks!

We’ll be reconvening in two weeks to discuss what else our WEF respected leaders have determined about the future of technology.

⚛️Addie LaMarr

Founder, LaMarr Labs

Status Report (The Intel)

Quantum-safe encryption ran on real internet cables and nothing broke

PacketLight and NEC ran quantum key distribution alongside live 400G telecom traffic without slowing it down. No dramatic failures while quantum-safe keys rode along as if they belonged there the whole time. Woo-hoo!

The Royal Navy took quantum navigation to the Arctic to see if it would survive

The Royal Navy tested GPS-free quantum navigation using cold atoms in a place that is cold, remote, and generally hostile to fragile technology. It worked. That’s the entire story, and it’s more impressive than it sounds.

Chile decided quantum belongs in the “real economy” category

Chile rolled out a ten-year quantum and biotech strategy, which is government-speak for “we plan to fund this and expect it to matter.” Quantum is slowly graduating from science project to infrastructure.

Michigan got $9M to build sensors we’ll pretend were obvious later

The University of Michigan picked up a $9 million grant for distributed quantum sensing. Right now it’s research. Later it’ll be something everyone assumes had to exist.

How it Works: Quantum Measurement

Think about a Snapchat you haven’t opened yet.

It could be a selfie, rant, or something you IMMEDIATELY regret seeing. 🥴

Right now, all of those possibilities are still alive.

The moment you tap it, reality snaps into place.

Now it’s one specific thing. Snapchat marks it as seen. The countdown starts. Whatever it was before you opened it is no longer available for inspection.

There’s no safe preview mode, “just looking,” and there’s def no rewind.

Opening the message is the whole event.

That’s exactly what quantum measurement is like.

When scientists measure a quantum system, they don’t gently check on it. They force it to pick an outcome. All the fragile information that existed beforehand disappears the moment it’s observed.

This is why quantum systems are so temperamental. It’s also why building a useful quantum computer feels like babysitting something that immediately panics when you look at it.

In the quantum world, observation is more like tapping the screen and accepting whatever you get.

Investment Watch

Not terribly much happened over the holiday season!

SEALSQ expands its Quantum Investment Fund to > $100M

SEALSQ Corp (NASDAQ: LAES) increased its dedicated Quantum Investment Fund from $35 million to over $100 million, aiming to accelerate Europe’s quantum-safe digital infrastructure and sovereign capabilities.

Field Notes

I keep running into this funny problem consulting on the quantum transition.

Everyone agrees cryptography is muy importante (very important). Very few people can tell you where theirs actually is.

And I don’t mean that in a “wow, security is broken” way. I mean in a very normal, very human way. Crypto was designed to be invisible. It did its job quietly. Nobody ever had to point at it and say, “this one, right here.”

So now when you ask, the answers get real vague, real fast.

I’ve been helping an aerospace company build a cryptographic inventory, and what surprised me is that the hard part isn’t super technical. Instead what’s difficult is the moment you realize you have to translate extremely ~nerdy~ parameters into something operational people can understand in about 10 minutes, preferably without their eyes glazing over.

Tech teams don’t always have this written down cleanly themselves. Operational folks don’t care about algorithms or key sizes anyway. They want to know what depends on what, what breaks if it fails, and who gets yelled at when it does. (anddddd, more importantly, whether they’ll be awarded better contracts in the future)

Then you realize you can’t even reuse the same explanation twice.

Healthcare lingo’s different from energy. Energy speaks differently than aerospace. Finance has its own anxieties and language. Tech has its own brand of chaos. You keep asking the same questions, but you keep having to ask them in a slightly different way.

I’ve been learning Spanish the past few years and it’s a surprisingly similar process.

Oddly enough, this is the part I like.

My ex-military background helps because the task feels familiar. You strip things down. You decide what matters now. You explain it plainly. You move forward without perfect information.

Most people don’t understand cryptography, and honestly, that was fine for a long time. The quantum transition is forcing it out into the open. Whether this goes smoothly or turns into a mess depends almost entirely on how good we get at explaining the invisible stuff.

Tangent

Fun Fact for 2026:

Quantum computers in 2026 still need to be colder than outer space, isolated from vibrations, and carefully babysat just to stay coherent for a few seconds.

Meanwhile, we’re already planning how to secure the global internet from them.

It’s funny because we’re redesigning cryptography to defend against a machine that currently can’t survive someone closing a door too hard.

The irony isn’t lost on me…

Hope you’ve all had an amazing holiday season!

⚛ Addie LaMarr

Founder, LaMarr Labs

P.S. what quantum concept do you want broken down next??? I read every reply!

I’ve been sitting on this for a veeeery long time.

If you’ve followed me on YouTube or Instagram, you know I’ve been talking about quantum threats forever. Fun fact: the very first video I ever uploaded to YouTube was a 1.5-hour seminar on quantum cryptography for ESTIN Algeria. I’m so excited that it’s finally relevant for everyone else.

As we head into 2026, the cryptographic transition is graduating from a “someday” thing to being mandatory across governments, critical infrastructure, and big companies that don’t want to wake up one day with broken security.

So this is THE moment we have been waiting for.

The Quantum S.H.I.F.T. is a bi-weekly briefing on what’s actually happening in quantum land including what’s real, what changed, and why you should care (and how to take advantage of this opportunity)

S.H.I.F.T. stands for:

1. Status Report: What changed & why it matters.

2. How it Works: One concept in plain English.

3. Investment Watch: Where money & power are moving.

4. Field Notes: Real-world lessons from my experience helping transitions.

5. Tangent: Anything else I want to discuss...

Let’s get into it.

Status Report (The Intel)

• Stanford demonstrated room-temperature entanglement on a nanoscale chip... without dilution refrigerators.

  They built a nanoscale chip that creates entangled particles without million-dollar freezers.

  Translation: removes a major deployment bottleneck for quantum networking research.

• Google quietly moved the RSA goalposts

  New estimates suggest 2048-bit RSA may be breakable with under ~1 million physical noisy qubits (down from 20 million) under optimistic error-correction assumptions.

  Translation: long-term risk timelines are compressing faster than expected.

• China vs. Google isn’t slowing down (updated in December 2025)

  China’s Zuchongzhi 3.0 is chasing raw speed. Google’s Willow is chasing error correction.

  Translation: different strategies, same pressure on classical public-key crypto.

How It Works

The Concept: Quantum Entanglement

If you’ve got kids, you already understand quantum entanglement via Elf on the Shelf.

Before December starts, Santa prepares two magical things at the same time:

• A Scout Elf for your house

• A matching Elf record that stays with Santa

They’re created together, which is why they’re linked (aka share the same state).

The Rule

You’re allowed to look at the Elf.

But if anyone touches it, moves it, or messes with it, the magic breaks.

There’s no way to fix it quietly.

The Classical Way

In a normal world, Santa would rely on messages.

Reports get sent.
Signals travel.
Messages can be intercepted or altered.

Trust depends on protecting the channel.

The Quantum SHIFT

With the Elf, there is no message.

The Elf isn’t reporting back.
Santa isn’t receiving updates.

Instead, both sides check their own record.

If the magic is intact, everything behaves exactly as expected.

If someone touched the Elf at any point?

The magic doesn’t line up anymore because the shared state was disturbed.

What That Means

Nothing has to travel between Santa and your house.

No information is transmitted this way... only correlations are revealed.

But interference can’t be hidden.

The system itself remembers it.

That’s entanglement.

Why This Is Weird (and Powerful)

In classical systems:

• You protect messages

• You assume observation’s harmless

In entangled systems:

• Observation changes the system

• Interference leaves fingerprints

Trust gets enforced by physics... not rules or software.

Entanglement means two things stop behaving like separate objects and start behaving like one system, even when they’re realllllllllllly far apart.

Investment Watch

• Quantinuum × Qatar’s $1B joint venture

  It’s a 10 year play. Not a funding round… more of a sovereignty move.

  Governments are buying control.

• Nu Quantum raised $60M

  Biggest Series A ever for quantum networking.

  Processors need pipes to scale... These are the pipes.

• Morgan Stanley took 7% of IonQ

  Wall Street officially cares now.

Field Notes (Where People Are Still Getting This Wrong)

One thing I keep seeing in the field is orgs (& people) treating quantum risk like a future mythical technical problem instead of the present risk management problem that it *CURRENTLY* is.

That distinction ~really~ matters.

From a risk perspective, you don’t wait for the event to happen.

You look at:

• Likelihood over time

• Impact when it happens

• Cost of mitigation now vs. later

Quantum fails all three tests if you delay.

1. The Threat Window’s Already Open

Risk isn’t about when quantum computers exist like many believe. Think of it more about how long your data needs to stay secure.

Adversaries are already harvesting encrypted traffic with the assumption that it will be decrypted later. If you’re protecting data with a confidentiality horizon of 5–10 years, that data is already exposed by today’s standards.

This is a classic deferred-risk problem.

2. The Impact Isn’t Incremental

When RSA fails, it’s not graceful or forgiving...

It’s a whole trust collapse:

• Certificates quit being trustworthy

• Digital signatures stop proving authenticity

• Secure channels quit being secure

• Identity systems stop being reliable

From a risk lens, this is pure systemic failure.

3. Delay Increases Both Cost & Blast Radius

Risk mitigation is cheapest when done early + deliberately.

Waiting compresses timelines, increases scope, and forces emergency decisions under regulatory and operational pressure. That’s when migrations get sloppy, expensive, and fragile.

Planned transitions are simple engineering projects, while unplanned transitions are total crisis response.

4. Early Action’s a Risk Reduction Strategy

Orgs that move early are reducing exposure…. by a LOT.

They:

• Shorten the harvest-now-decrypt-later window

• Spread migration costs over time

• Avoid compliance pileups

• Maintain operational control

That’s textbook risk management.

Bottom line:

Waiting compounds risk into something unmanageable.

From a risk standpoint, the safest move is to start the SHIFT while you still have choices.

Tangent

2025 is the 100 year anniversary of quantum mechanics.

In 1925, Werner Heisenberg went to a remote island because his allergies were ruining his life. While attempting to breathe normally, he came up with the math that’d eventually lead to quantum mechanics… and, a century later, threaten the internet as we know it.

There’s a lesson there (and it isn’t to forego allergy meds to gain Megamind status... though I might try this Spring)

Big shifts don’t always have to start with grand plans. Sometimes they start with someone annoyed, uncomfy, and trying to fix a problem the best way they can.

Also… until you open that gift under the tree, it is simultaneously exactly what you wanted & definitely socks.

Quantum says both are true.

Observation ruins EVERYTHING :P

Happy Holidays.

Enjoy the last few calm days of the classical world.

We start the SHIFT for real in January.

Addie LaMarr