Guided tour
11 short steps. Each one ends on a real page with the example already loaded, and tells you what you should see — because on this subject the intuitive answer is usually the wrong one, and being wrong once teaches more than three paragraphs explaining why. Start anywhere; nothing is stored about you.
Foundations
- 1. A qubit is not both 0 and 1The usual explanation of superposition is misleading, and the Workbench can show you why in one circuit.
- 2. Correlation is what entanglement looks likeA Bell pair and two independent coin flips are identical in a histogram. The difference only shows in the correlations.
- 3. Depth is the number that costs youTwo circuits with the same gate count can take very different times to run and decohere by very different amounts.
Error correction
- 4. A code never looks at what it protectsError correction has to detect errors without measuring the data, because measuring it would destroy the superposition.
- 5. Distance 3 corrects one error, and breaks on twoThe relationship between code distance and correctable errors is the most commonly misread number in the field.
- 6. The failure a syndrome cannot showSome errors leave no trace at all, and that is precisely what a code's distance measures.
- 7. Above threshold, more qubits make it worseError correction is not a dial you turn up. Below the threshold it works and above it actively harms.
- 8. Most of a fault-tolerant computer makes magic statesA resource estimate quoted as a qubit count is usually quoting the smallest part of the machine.
Evidence
- 9. Zero observed failures is not a rate of zeroThe most persuasive number on a leaderboard is often the one with the least evidence behind it.
- 10. Two numbers are not a comparisonMost invalid scientific comparisons look exactly like valid ones until you check what was held fixed.
- 11. A matching hash does not prove where a number came fromReproducibility hashing is load-bearing here, and it is routinely asked to prove something it cannot.