The public record of every prediction
A prediction goes on this page the day it is issued, in writing and with its date, and stays here however it turns out, in the same type size either way. Later notes go underneath it, with a date. Each one is about how narratives move, never about how prices move.
The Record’s own error rate appears here once 50 predictions have resolved. That number is proposed, and will be set before the first prediction.
Predictions open and resolved
None so far. A new prediction appears here the day it is issued, in the ten fields set out in the Method. Once its date arrives it shows the raw count, the result, and its squared error beside that of the trivial prediction. The register below is a summary, and the ten fields sit with each entry.
As of November 5, 2026: 0 issued, 0 open, 0 resolved, 0 void, 1 declined, 0 corrections.
Resolved0An error rate needs 50
Proposed
No prediction has been issued.All 50 rows are blank.
| No. | Issued | Claim | Likelihood | Resolves | Result | Raw count | Squared error | Trivial error | Later notes |
|---|---|---|---|---|---|---|---|---|---|
| 01 | |||||||||
| 02 | |||||||||
| 03 | |||||||||
| 04 | |||||||||
| 05 | |||||||||
| 06 | |||||||||
| 07 | |||||||||
| 08 | |||||||||
| 09 | |||||||||
| 10 | |||||||||
| 11 | |||||||||
| 12 | |||||||||
| 13 | |||||||||
| 14 | |||||||||
| 15 | |||||||||
| 16 | |||||||||
| 17 | |||||||||
| 18 | |||||||||
| 19 | |||||||||
| 20 | |||||||||
| 21 | |||||||||
| 22 | |||||||||
| 23 | |||||||||
| 24 | |||||||||
| 25 | |||||||||
| 26 | |||||||||
| 27 | |||||||||
| 28 | |||||||||
| 29 | |||||||||
| 30 | |||||||||
| 31 | |||||||||
| 32 | |||||||||
| 33 | |||||||||
| 34 | |||||||||
| 35 | |||||||||
| 36 | |||||||||
| 37 | |||||||||
| 38 | |||||||||
| 39 | |||||||||
| 40 | |||||||||
| 41 | |||||||||
| 42 | |||||||||
| 43 | |||||||||
| 44 | |||||||||
| 45 | |||||||||
| 46 | |||||||||
| 47 | |||||||||
| 48 | |||||||||
| 49 | |||||||||
| 50 |
13 to 50are blank too.
Own error rateNoneAppears once 50 have resolved.
The prediction that was declined
Draft, declined under the Method on November 5, 2026 and never issued.
Will the Federal Register publish at least one new document containing the words bell to bell by February 3, 2027?
Not product output. AI agents drafted it.
Error rates are not yet published.
It comes from the first file.
- Narrative
- The label bell to bell, used in school and state text, has appeared once in federal text.
- Series
- Search for “bell-to-bell” and “bell to bell” in the Federal Register public document search, counted as distinct document numbers. One document on November 5, 2026: 2026-16590, published August 14, 2026.
- Target
- At least one Federal Register document published from November 6, 2026 through February 3, 2027 is returned for either spelling.
- Dates
- Issued: None. Would resolve February 4, 2027.
- Rule
- On February 4, 2027, run the two searches below in any browser. They search the Federal Register for each spelling, limited to the window.
- List the document numbers returned. Count each number once, whatever the document is about.
- If the count is 1 or more, the result is Yes. If it is 0, the result is No.
- If either search still fails on February 11, 2027, the result is Void.
Search for bell-to-bell (raw result)Search for bell to bell (raw result)
- Likelihood
- None. A declined prediction shows no likelihood.
- Confidence
- None, for the same reason. The series is one document long.
- What would change this
- A new FCC document in the E-Rate proceeding (WC Docket No. 26-133) that repeats the words. A federal notice on school phone rules from any agency. Reply comments there close November 12, 2026, and no next step in that proceeding was read.
- Trivial prediction
- The last 90 days held one document. Repeating the last window is a mechanical baseline, and it says Yes. SOVRINT makes no prediction here.
- Author, who paid, how it was made
- AI agents drafted it, and no client paid for it.
- Scoring
- None. A declined prediction is left unscored.
- Rule tested
- Both searches were run with the rule on November 5, 2026. The window ahead returned 0. Moved back 90 days, to August 8 through November 5, 2026, each returned 2026-16590 alone, and the rule would have said Yes.
The Method declines a prediction when its error is unmeasured, its series is short, or its publishers are too few. This one meets all three and stands on the Record as declined. Its label makes no claim either: the entry under Narrative describes a label and falls short of a narrative under section 1.1. It stays here as a worked example of what a prediction looks like. Issuing it would take a written change to the Method first. Next review: February 4, 2027, against what would change this.
Note of November 5, 2026: Trends has since counted this search by week back to 2016. The decline stands on the unmeasured error and the single publisher.
Corrections
None so far. An error of fact gets an entry here saying what the page said, what it says now, the date, and who reported it if they agree to be named. The wrong text stays on its page, struck through. To report one, write through Contact and choose A correction.