The conference talk, and the three-minute version

1 · Learn the move · Enhance the Output

A talk is an output-specification problem: twelve minutes is roughly ten to twelve slides, one message per slide stated in the slide title as a sentence, timing marks every three slides so drift gets caught at minute four rather than minute eleven. Data slides carry their own defensibility - the n, the test, the error-bar definition, on the slide, because the person who will ask is in the second row. The three-minute version is built by cutting the twelve, not by writing fresh - same claim, two supports instead of six, one figure. If the twelve-minute talk cannot be cut to three, its structure was decorative, and the cut list is the diagnostic: every slide is marked KEEP (the claim dies without it) or CUT (supporting texture), and the KEEPs must connect on their own.

Build my talk from these results: [paste figures + claims with n/tests]. Output spec: [12] minutes, one message per slide as a sentence title, timing marks every 3 slides, n + test + error-bar definition on every data slide. Then produce the 3-minute version by cutting: mark every slide KEEP or CUT with one line of why, and verify the KEEPs still connect as an argument.