Traced claims
A sentence that is circulating, the rows in this archive that carry it, the earliest of those rows, and what the named parties published on the same subject. These pages state no conclusion about any claim.
What is traced
1 claim. Sorted by the date of the earliest instance this archive holds, which is arithmetic. Nothing here is ordered by how well sourced anything is.
| Claim | Subject | Sources | Earliest row published |
|---|---|---|---|
| Memory is undersupplied, and the shortage runs into 2027. | DRAM and NAND supply | 4 | 2026-08-11 |
How a claim gets in
A claim enters this register on repetition, not on suspicion: it must appear in 4 or more distinct collected sources inside one 30-day window. The threshold and the window are set once and printed on the page.
Selection is editorial even when the prose is not. Entries chosen because somebody suspected a claim was weak would carry a thesis with no adjectives in it, which is the opinion layer wearing a lab coat. A mechanical threshold does not depend on anyone's view of whether the claim is true.
Why this number. Four, set 25 August 2026. The archive collected its first item on 13 August 2026, so a 30-day window today reaches 12 days of collection rather than 30. Five would have admitted nothing. This number is a decision and not a measurement, it is printed for that reason, and raising it later drops entries that are already published.
A trace that comes back impeccably sourced stays in the register. Those entries are the proof it is not a list of accusations, and dropping one because it came back clean is the exact failure the rule exists to prevent.
What these pages do not do
It states no verdict. Not weakly sourced, not unverified, not misleading, not confirmed. The pages print the trace and the reader draws the conclusion. Nothing here may sort, count or badge by how well sourced a claim is, because that is a score and this record does not score.
What this archive holds. Titles, publishers, dates and links. Not article bodies. So the question of what an item itself cited cannot be answered by any query here: somebody has to open the link and read it. Every entry records whether that reading happened, and where it did not, it says so rather than leaving the field empty.
Candidates are mechanical. Noticing that several items are making the same claim is human, and no query does it. Every entry names who added it and when.