SymbiQ is written and kept by one person, a stubborn test suite, and no money or ties to any company whose claims turn up on these pages.
The independence matters because most of what happens here is arguing with numbers. A figure you read last week turns out to be wrong, or right, or right about a different quantity than the headline implied. Anyone doing that from behind a curtain is asking for trust on credit. So instead of a biography, this page lays out the machinery, and every part of it can be checked from the outside.
How AI is used here
Large parts of this site are drafted by AI, and saying so plainly is part of the method. Three things do the checking, and each has limits, set out in full on How we check.
AI drafts. Most of the prose, the code, the calculators and figures, the Ledger write-ups and the proposed verdicts on it, and the question bank.
Scripts check. They recompute arithmetic and game pars, read counts from the pages they describe, look every arXiv citation up against arXiv, and compare tables with the sentences around them.
Fresh AI sessions attack. A referee pass with no memory of writing the page recomputes the numbers and fetches the sources. It can share the drafting AI’s blind spots.
A person decides. The admin chooses what is published, reads the headline claims and the sources behind them, and acts on reader reports. Not every proof is re-derived every proof line by line.
How a number gets onto this site
Figures that carry weight come from the primary source: the paper, the standard, the regulation, the press release itself. A confident summary by somebody else doesn’t count. When a number is calculated rather than quoted, a script recomputes it. Where you might reasonably want to see the working, the page shows it in a small receipt that opens beside the claim.
Three kinds of check run against the site:
Arithmetic gets redone. Widget outputs, thresholds, game pars and resource estimates are recomputed by scripts, usually by trying every possible input instead of a hopeful handful. A game par that isn’t a proven optimum doesn’t ship.
Citations get checked for rot. Papers age like milk more often than wine. A script looks up every arXiv identifier the site cites against arXiv itself: does the paper still exist, was it withdrawn, has it moved several versions past the figure we quote, and does our sentence name the right authors or title? The worst mistake this site has made was a stale number from a superseded paper, which is why this guard exists.
Sentences get checked against data. The nastiest bugs here were never sums. They were sentences that confidently contradicted a table sitting three inches below them, so some of the checks now read the prose around a number too.
Claims are judged by rules fixed before the verdict is written
The Ledger tracks public, dated, falsifiable claims about quantum computing made by labs, executives and regulators. Its rules are the interesting part:
The resolution criteria, and the exact sources that will settle them, are frozen the day a claim is logged. For most claims that is long before the deadline; two were logged after their deadlines had passed, and their entries say so. Moving the goalposts later removes the claim from scoring and leaves a public note explaining why.
A verdict is written after the deadline, in its own pass. It is checked against the criteria and sources frozen when the claim was logged, and it cites every piece of evidence it uses.
The verdict vocabulary has no word for lying, hype or fraud. The Ledger sets what was said beside what happened and leaves the adjectives to you.
If a claim turns out to be unscoreable, that is our failure. It goes on our scorecard, as visibly as anyone else’s miss.
What happens when this site is wrong
It goes on the corrections log, in public, with more room than the original mistake got: what we said, what is true, and how we missed it. The log counts anything wrong that reached the published site, whoever spotted it, including errors we caught ourselves and ones nobody else would have noticed. Nothing gets painted over.
An error you find is the most useful thing you can send us, and a reader-found correction runs under your name.
What this site is not
Not financial advice, ever. Several pages discuss cryptography, Bitcoin and listed companies. None of it is a recommendation to buy or sell, and no position is held in any company named on this site.
Not sponsored, and not affiliated. No vendor pays for coverage, sees copy before it runs, or gets a say in a Ledger verdict. There are no ads.
Not a consensus view. Where experts disagree, the page names the disagreement and gives both sides their best argument.
Not finished. Some sections are incomplete, and those sections say so on the page.
Telling us we got something wrong
The fastest route is the correction report. It lands on the desk directly, and it is read by the desk.