Last week I predicted, in public and in advance, how coverage of the September 24 Xi summit would split before anyone knew what would be signed, said, or shown. That piece made one claim worth isolating from everything else in it: the frame is often set before the facts are. Not because reporters are dishonest, but because two audiences have already decided, months in advance, what kind of story they need an event to tell — and the event, whatever actually happens at it, gets sorted into that story afterward.
That claim is either a real method or it's a parlor trick that only works once, on one summit, in hindsight-adjacent conditions. The way to find out is to generalize it, state the generalization as a testable rule, and then apply it again — live, dated, before the outcome is known — to something that has nothing to do with China.
The Rule
Media frames are predictable in advance whenever three conditions hold: the event is recurring or anticipated, so outlets have already built the interpretive scaffolding they'll hang the next instance on; the participating institutions have stable, known incentives, so which facts get selected is a function of who's speaking, not what happened; and the underlying reality is genuinely mixed — complex enough that multiple true descriptions are available, so no outlet has to invent anything to reach its preferred conclusion.
When those three conditions hold, you are not predicting the future. You are predicting the selection. The event can go any number of ways and the frame will barely move, because the frame was never a response to the event in the first place — it was a response to what each audience already needed to believe.
The Test That Separates This From Punditry
Anyone can say “the media will spin this” after the fact. That costs nothing and proves nothing. A predictive-framing claim only counts if it clears three bars, and I'd hold myself to all three going forward:
— Dated. Published before the event, with a timestamp that can't be quietly edited later.
— Specific. Naming which outlets or outlet-types will lead with which facts — not “some will spin it,” but which frame goes with which incentive.
— Falsifiable. Stated in a way that a wrong guess would actually look wrong. If the prediction is vague enough to fit whatever happens, it isn't a prediction.
The Xi piece cleared all three: dated July 29, naming the dealmaker frame and the predatory-strategy frame specifically, checkable against the bylines that run on September 25. Here is a second one, cleared the same way, resolving sooner.
Case Two: The August 7 Jobs Report
At 8:30 a.m. Eastern on August 7, the Bureau of Labor Statistics releases the July employment report — nonfarm payrolls, the unemployment rate, wage growth — the last major data point the Federal Reserve sees before its September meeting. Whatever the number is, the coverage is already sortable into four frames, and here they are, named before the release:
“Record job growth” · “Labor market cooling” · “Inflation risk rising” · “Wage stagnation continues”
All four are available almost regardless of the topline number, because a single jobs report contains enough true, simultaneous facts — the headline payroll figure, the unemployment rate, wage growth, sector-level detail, revisions to the prior two months — to support any of them without anyone needing to misstate a thing. An outlet inclined to emphasize labor-market strength leads with payrolls and buries the wage-growth softness in paragraph nine. An outlet inclined toward concern leads with the revision to the prior month and treats the headline number as provisional. Both pieces will be accurate. Neither will be the same story.
The specific prediction: outlets already positioned as favorable toward the administration's economic record will lead with the payroll or wage figure that reads strongest and treat any softness as noise; outlets positioned skeptically will lead with revisions or the unemployment rate and treat the headline as the noise. This is checkable on August 7, not in retrospect — and because the release feeds directly into September Fed expectations, watch particularly for which outlets frame the number as ‘supporting a rate cut’ versus ‘complicating the case for one.’ Same data, and the monetary-policy conclusion will split along the same line as the labor-market conclusion, because it's the same selection running twice.
Why the Taxonomy Is Long but the Method Is Short
The list of places this applies is genuinely large — recurring summits, Supreme Court rulings, major legislation, natural disasters, high-profile trials, corporate and regulatory fights. It's tempting to treat that length as the finding. It isn't. The finding is narrow: wherever an event is anticipated, the participants' incentives are known, and the reality is mixed enough to support more than one true account, the frame can be named in advance. The length of the list is just a symptom of how often those three conditions co-occur in a modern news cycle — which is to say, constantly.
That's also the honest limit of the method, and it's worth stating plainly rather than overselling: predictive framing works on the recurring and the institutional. It does not work on the genuinely novel — an unprecedented event, a true black swan, has no scaffolding to slot into yet, and the first coverage of something truly new is often more honestly confused than the coverage of anything on this list. The method is powerful precisely because so much of what gets covered as news is not novel. It's a rerun of a category with a new date attached, and the category is what the audience already knows how to feel about.
What This Is For
Not cynicism. The point of naming the frame in advance isn't to conclude that nothing can be known or that every outlet is equally compromised — that's the post-truth adaptation this whole project has been arguing against from the start. The point is narrower and more useful: a reader who can predict the frame before the story runs is a reader who reads past it, because the frame stops being invisible. You can't see architecture you're standing inside of, until you've watched it get built in advance, on the record, and checked the blueprint against the building.
Two predictions are now on the record: September 25 for the summit, August 7 for the jobs report. Both dated. Both specific. Both checkable by anyone, including me.