The midterms, and what this site will not say about them

Election day is Tuesday 3 November 2026. Between now and then there will be no shortage of people willing to tell you what it all means.

This site will not do that, and it is worth saying plainly what it will do instead, because the difference is easy to mistake for timidity.

What will not appear here

No forecast. No probability. No projected seat count, no recommended position, no market odds, no ranking of who is ahead. Not as a headline, not in a chart, not hedged in a paragraph at the bottom.

That is not a preference about tone. It is enforced in the code: the publishing path refuses a draft carrying forecast language, and the tests that hold it there are the ones nobody is allowed to weaken to ship a feature. A number about the future cannot be checked against a record, and a system whose entire claim is "every figure has a source you can open" cannot also print figures that have none.

There is a quieter reason as well. A forecast is the one thing a reader will remember and repeat, and it is the one thing this pipeline could never stand behind afterwards. Being right about it would be luck. Being wrong about it would be the only thing anyone recalls.

What will appear here

Claims made in public, set against the primary record, with a date.

Candidates and incumbents make an enormous number of statements that are not predictions at all. They are assertions about what happened: who filed a bill, who voted which way, what a programme cost, which counties a disaster declaration covered, how many executive orders were signed in a period. Every one of those has a record behind it, and the record does not care who is ahead.

Those are the claims this site can settle, and an election season is when they are made in volume.

Two dates worth watching

1 October 2026. The federal fiscal year turns over. Appropriations claims cluster around it, and spending assertions are among the few things that can be answered to the dollar from a published record rather than argued about.

3 November 2026. Election day itself. Nothing this site publishes will change on that day, which is the point: a claim about the record resolves the same way the week before an election and the week after it.

If we get one wrong

We publish the correction beside the original and supersede it rather than editing it away. Two answers have been corrected so far and both remain visible, with what they said, when they said it, and what replaced them. That is the only promise here that actually matters in an election season, because it is the one that costs something.

The record is the published set. The pipeline is everything else, including what we could not settle.