Corrections

When something on the site is wrong, it's fixed and recorded here, so you can see what changed and when.

What gets corrected, and how

• Errors of fact, such as a wrong date, figure, attribution or description of a standard, are fixed as soon as they're confirmed, and listed below.
• Clarifications that change what a reader would take away are listed too.
• Small edits, such as typos, wording and layout, are made without a log entry.
• Each page shows the date it was last reviewed.

To report an error, email hello@plainquantum.com with the page and, if you can, a public source. How sources are chosen is set out in How this site is made.

Corrections log

Newest first. The log starts on 12 October 2026.

• 19 October 2026, How a quantum algorithm works, What is Shor's algorithm?, What is Grover's algorithm? and What is quantum advantage? Corrections from an independent AI scientific check. How a quantum algorithm works presents spread, mark and mix as one useful pattern, used especially by search algorithms such as Grover's, rather than the shape most algorithms share. What is Shor's algorithm? no longer says that measuring reveals the period directly: ordinary calculations interpret the measurement results, and the procedure can need more than one attempt. What is Grover's algorithm? says that attacking real RSA keys with Shor's algorithm would need a large, reliable quantum computer that has not been publicly demonstrated. What is quantum advantage? describes a fair comparison with strong known classical methods for the same task and accuracy, and no longer suggests that error-corrected qubits are needed for every advantage demonstration.
• 19 October 2026, What is a logical qubit? and How quantum gates work. Corrections from an independent AI scientific check. What is a logical qubit? no longer says that error-correction checks always reveal where an error happened: the pattern points to the most likely error and doesn't always pin down a single faulty qubit. It also no longer says that every useful algorithm is measured in logical qubits: many long, demanding algorithms are likely to need them, while shorter experiments can use physical qubits directly. Its threshold sentence now says that below the relevant threshold, increasing the code size can reduce logical errors, instead of claiming that above it adding qubits always makes things worse. How quantum gates work now explains reversibility as ideal gates preserving quantum information, and notes that measurements and resets are different kinds of operations.
• 19 October 2026, What is quantum computing?, What quantum computers are good at, Why qubits are so fragile and How to read a quantum headline. Corrections from an independent AI scientific check. What is quantum computing? and What quantum computers are good at no longer say that a quantum computer helps only when wrong answers cancel: simulating quantum systems is another use, and neither speeds up every problem. What quantum computers are good at no longer says a million numbers fit in twenty qubits, and says that a measurement returns one result per run, not the list of amplitudes. Why qubits are so fragile no longer describes decoherence as literally an accidental measurement, and says that error-correction checks help infer likely errors rather than always pointing to the qubit that failed. How to read a quantum headline no longer says that every useful algorithm needs logical qubits, and asks for a fair comparison with the strongest known classical methods.
• 19 October 2026, Post-quantum cryptography for non-engineers. Two new figures, checked by an independent AI scientific review: what a quantum computer would break and what replaces it, and example key and signature sizes. The hybrid-handshake caption now separates agreeing the session key from proving who the server is, which uses the private key matching the server's certificate and is mostly not post-quantum yet.
• 19 October 2026, What is a qubit?, Superposition, Entanglement and How a quantum algorithm works. Corrections from an independent AI scientific check of the Recall questions. Superposition no longer says that repeated measurement reveals the amplitudes: it shows the chance of each result, and signs and phases need measurements in several different ways (What is a qubit? now says the same). Entanglement no longer says that separate qubits can't always match: classically correlated qubits can, and what only entanglement does shows up across different measurements. What is a qubit? qualifies the spinning-coin analogy, Superposition names the Hadamard (H) gate in its worked example, and How a quantum algorithm works notes that Grover's search overshoots if it runs too many rounds.
• 19 October 2026, Method, the readiness self-assessment and Privacy. Method and the self-assessment said the site uses cookie-free page-view analytics before that had been checked against the hosting configuration. It has now been checked: Cloudflare Web Analytics is switched on. Both pages now point to the Privacy page, which says exactly what it records. Method also gains a section on how AI is used and who is accountable.
• 17 October 2026, banking guide chapter 8 and Auditing a post-quantum programme. Chapter 8 now limits NIST's proposed 2030 step to the 112-bit security level and notes that neither NIST date is an OSFI deadline. The audit article now follows the IIA's 2026 Three Lines Statement of Position: the board sets risk appetite (the article had said second-line risk did), internal audit has a functional reporting line to the board, and the IIA's safeguards apply if internal audit takes on second-line work. Its description of a completed migration no longer says the old algorithms must be gone: a properly enforced hybrid, isolation or tunnelling can also mitigate the quantum risk.
• 17 October 2026, consistency update across the learning path, quick answers and Canada page. Pages that described Craig Gidney's 2025 estimate as the latest now also cite the 2026 Pinnacle designs (about 94,000 to 120,000 physical qubits, theoretical, using newer error-correcting codes). Credited the 2019 estimate to Craig Gidney and Martin Ekerå. "Post-quantum in Canada" now cites the Treasury Board's Security Policy Implementation Notice as the source of the binding federal dates, and notes OSFI's statement that its technology risk bulletins are not regulatory expectations. Reworded two pages that said OSFI "expects" quantum risk to be managed under B-13; they now say the guideline still applies. Banking guide chapter 7: the regulatory table now cites the Treasury Board notice as the binding authority for the federal dates, limits NIST IR 8547's proposed 2030 step to the 112-bit security level, and distinguishes B-13, which applies today, from OSFI's non-binding quantum-readiness bulletin. The CRQC quick answer now presents the 2026 figures as theoretical RSA-2048 estimates, not a general threshold. Three more pages that summarised NIST's proposed 2030 step as covering all RSA and elliptic-curve algorithms now limit it to the 112-bit level.
• 14 October 2026, learning path and Why qubits are so fragile. Updated the number of logical qubits demonstrated so far (dozens, not "a handful") and the size of entangled states (close to a hundred), citing Quantinuum's 2025 Helios results. Clarified that hybrid key exchange stops recorded traffic being decrypted later, not being recorded. Also a clarity pass: links to quick answers, a reordered worked example, less repetition.
• 13 October 2026, learning path and home-page demo. The qubit demo now clears its counts when you move the slider, so the tally always matches the weighting shown. Made the 2024 error-correction milestone more precise (first below-threshold demonstration, with Google's 2023 result now cited), limited a CNOT statement to gate-based machines, and reworded the superposition analogy.
• 13 October 2026, Post-quantum cryptography for non-engineers. Replaced a 2024 traffic figure with Cloudflare's 2025 figure, noted that hybrid TLS is now an internet standard (RFC 10024), renamed "Deadlines to know" to "Dates to know", and scoped the CNSA 2.0 and EU dates. Small wording fixes elsewhere so that policy dates aren't described as bank deadlines.
• 12 October 2026, banking guide chapters 1 to 3, the inventory template and the Q-day quick answer. Renamed the quantum status labels from "broken, weakened, fine" to "vulnerable, check key size, not affected", and defined them in chapter 1, so they no longer imply anything is broken today.
• 12 October 2026, Banking guide chapter 1. Clarified that working out a private key from a public key needs a sufficiently capable quantum computer, and applies to RSA and elliptic-curve cryptography.
• 12 October 2026, Why your bank is already preparing; Banking guide chapter 7. Clarified that the commonly cited 2030 and 2035 dates come from government roadmaps and draft guidance, not deadlines set for banks.
• 12 October 2026, four quick answers. On phones, the comparison tables on the ML-KEM, CRQC, logical qubit and QKD pages briefly showed column labels shifted by one. Fixed the same day.
• 12 October 2026, companion CBOM project. Project Leap figures were corrected to match the BIS report, an unverified SWIFT date was removed, and the project page now states up front that it's an illustrative reference model.