f2 comes from the interviews: for each stack identity, how to frame a proposition so THIS voter engages and considers it - their language, their hierarchy, their moveable layers. f1 comes from the context store: how to frame what we are selling - the candidate, rendered through her record and proposition set. Neither alone is a message. The synthesis of both is.
f2 - FRAMING (from the READ) voter side
Stack identity B, moveable layers 1-2: "doing everything right, still falling behind""the math of a family doesn't work anymore"
Suggested narrative segments - candidate-free frames in the voter's own belief language. Fixed layers contribute constraints only: affirm, route around, stay silent.
f1 - SELLING (from the context store) candidate side
Hines, framed for this frame: "Warren mom of three who knows that math""prosecutor - held powerful people accountable"
Narrative suffixes - what we are selling, pre-shaped to attach to a frame. The same candidate yields different suffixes for different frames.
Operator gate #1: narrative segment suggestions are approved at the segment level before anything downstream runs.
The Synthesis
Frame + Sale = a Synthesized Proposition
The synthesis step merges the two into a new proposition: the f2 frame is the prefix - meet the voter inside their own belief - and the f1 suffix is the sale, attached where the frame naturally leads. Then the game shapes the ask. Same synthesis, different game, different creative.
FRAME: "doing everything right, still falling behind"+SALE: "Warren mom of three who knows that math"= SYNTHESIZED PROPOSITION: "The people doing everything right deserve someone who knows their math - and she lives it."
SIMULATION
You did everything right. The math still doesn't work.
Christina Hines raises three kids in Warren on the same math you do. Lower the cost of essentials. Hold the price-gougers accountable.
game: REINFORCE-MOB · ask = commit · same synthesized proposition
SIMULATION
Doing everything right? Then do the one thing that counts.
Early voting is open. The Warren mom who knows your math is on the ballot.
game: ACTIVATION-MOB · ask = turn out · same synthesized proposition
Operator gate #2: every rendered piece passes approve / edit / regenerate before deployment. The generator is opposed by a checker holding the full context store - claims outside the record do not survive the check. Guardrails are structural, not policy: fixed layers are not renderable targets.
The Movement
How narratives move a stack - salience, not conversion
The claim, precisely: repeated exposure to a synthesized proposition framed in the voter's own belief language raises the salience of the favorable moveable layer. The stack reorders. And because the sale is attached to that layer, alignment with the candidate rises as a consequence of the reorder - no belief was changed, no identity challenged. Fixed layers are load-bearing walls: never pushed, never touched.
STACK A VOTER - BEFORE ILLUSTRATIVE
1Law & orderMOVEABLE
2Cost of essentialsMOVEABLE
3Government corruptionFIXED
4Wages & jobsFIXED
WHAT HAPPENED mechanism
Press play. Weeks of cost-of-essentials framing - in this voter's language, attached to a candidate who fits - and the salience of layer 2 climbs past layer 1. Nothing was argued. Priority moved.
The proof pattern exists at city scale: the 2025 NYC race - everyone sold public safety at position one; one campaign found affordability sitting at position two or three across the whole city and promoted it. No minds changed. Priorities reordered. Elections are won on salience.
Claims boundary: movement is claimed on moveable layers only, measured two ways - engagement differentials on framed vs unframed creative (pixel-resolved), and re-interview panels watching the ordering itself move in the voter's own words. Post-race, results grade every claim. Reordering shown is ILLUSTRATIVE of the mechanism.
The Loop, End to End
One picture
READ → stack identities, moveable layers →f2: narrative segment suggestions (operator approves) Context store → record + proposition set →f1: narrative suffixes - how to frame what we are selling SYNTHESIS → Synthesized Propositions→ shaped per game→ creative engine (checker + approval queue) Deploy per flight → signal returns → salience reorders, board mass moves → reprice, re-select, regenerate on f2 shifts
Perfectly framing what we are selling - per stack identity, per game, at flight resolution - with a human gate at every consequential step. That is the Narrative Engine.