Everyone tells you to cut words from your slides. Nobody tells you the number.
Patrick Winston did. He told the same room the same numbers every January for forty years.
You commented DECK, so here are the four prompts that finish the talk, then the three from the reel, then the part that would not fit.
Who he was
Ford Professor of Artificial Intelligence at MIT. He ran the AI Lab from 1972 to 1997. Every January, during MIT's Independent Activities Period, he gave a one hour lecture called "How to Speak" to an overflowing room. He did that for over forty years. He died in July 2019, aged 76.
MIT put the recording and the full transcript online, free: ocw.mit.edu/courses/res-tll-005-how-to-speak-january-iap-2018
Every quote below is his, from that transcript.
The four prompts that finish the talk
4. The middle: say it three times without sounding like it
ROLE: You are structuring the body of my talk using Winston's cycling and verbal punctuation.
TASK: Make my three main points survive an audience that is 20% absent at any moment.
STEPS: Take my main points. For each, give me three passes: the plain statement, the worked example, and the callback later in the talk. Then place a verbal punctuation marker at each seam, an explicit "that was two of three, here is the third", so anyone who drifted knows where to rejoin. Then give me one question to ask the room at the halfway mark.
RULES: Cycling is not repeating the same sentence. Each pass has to earn its place. The question must be answerable by someone who was half listening, and not so obvious that answering it is embarrassing. After asking it, I hold silence for seven seconds. Write "[hold 7s]" into the script so I actually do.
OUTPUT: The body in order, the three passes per point, the seam lines, and the question with the hold marked.5. The evidence: build a fence around the idea
ROLE: You are applying Winston's fence: separating my idea from the one it will be confused with.
TASK: Stop the room filing my work under somebody else's.
STEPS: Name the two or three existing ideas mine will be mistaken for. For each, write one sentence of the form "this is similar to X, except". Then find my near miss: the closest thing to my idea that is NOT my idea, and explain what makes it fall short.
RULES: The fence goes in the middle of the talk, not the beginning, and it is one sentence per neighbour. It is not a literature review. If I cannot say what mine does that the neighbour cannot, I do not have a contribution yet, and you should tell me that plainly.
OUTPUT: The neighbours, one fence sentence each, the near miss, and a flag if the fence is doing the work my contribution should be doing.6. The close: the last slide and the last words
ROLE: You are ending my talk Winston's way.
TASK: Replace whatever my last slide currently is.
STEPS: Build a final slide titled Contributions, listing what I did, not what I concluded. Move every collaborator name to slide one. Then write my final words with no "thank you" in them. Give me two options: a salute to the room, something specific about this audience and this place, and a closing joke.
RULES: Kill any slide titled Conclusions, Questions, Thank You, or a bare URL. That slide sits up for twenty minutes while people ask questions and file out, so it is the most valuable real estate in the deck. A joke works at the END, never at the start, because by then they have adjusted to my voice.
OUTPUT: The contributions slide, the salute, the joke option, and the exact final sentence.7. The room: the part nobody edits
ROLE: You are checking the conditions I am presenting in, not the deck.
TASK: Find what will sink this talk before I open my mouth.
STEPS: Ask me the time of day, the room, the lighting, the seat count, whether it is recorded, and what happens immediately before my slot. Flag every risk. Then tell me which of my slides fail if the room is bright, and which fail if the back row is thirty feet away.
RULES: Winston's targets: 11am, because the room is awake and not yet fed, and lights FULL UP, because a dark room tells a human body to sleep. His line: "it is extremely hard to see slides through closed eyelids." An oversubscribed small room beats a half empty big one. If I am on straight after lunch, say so and tell me what to change.
OUTPUT: A risk list with the fix next to each, and the slides that need rebuilding for the actual room."It is a lecture about overhead projectors"
It is not, and here is why the rules survive the technology.
You have one language processor. Winston: "we have only one language processor, and we can either use it to read stuff or to listen to the speaker." That is a fact about the audience, not about the projector.
A student of his ran the experiment. Half the material spoken, half on the slides, then reversed for a control group. What people remembered was what they read. One subject wrote in the after action report: "I wish you hadn't talked so much. It was distracting."
About 20% of the room is gone at any moment, by his estimate, however good the talk is. Every structural rule he gives exists to let those people back in.
What did not fit in the reel
1. The formula. He opens by telling the room their success will be determined by their ability to speak, their ability to write, and the quality of their ideas, in that order. Then the formula behind it: knowledge, times how much you practise with it, times inherent talent. On the talent term: "notice that the T is very small."
2. Boards, props and slides are not interchangeable. Slides expose ideas. Boards teach them, because the speed of your hand is roughly the speed an audience absorbs at. Props get remembered for decades: someone recognised him twenty years on as the one who broke the pointer.
3. The seven second pause. Ask the room a question and you can hold silence for seven seconds. He counted it. It feels like an eternity from the front, and it is the standard. Pitch the question carefully: too obvious and nobody answers out of embarrassment, too hard and nobody can.
4. "Conclusions" loses to "Contributions". His words: "these are perfectly legitimate conclusions that nobody cares about. What they care about is what you have done." And the collaborators slide belongs at the front, not the end. A wall of names on the last slide reads as: divide this work by that many people.
5. Why "thank you" is a weak ending. It "suggests that everybody has stayed that long out of politeness and that they had a profound desire to be somewhere else." Once the applause starts you can mouth it. It is just not the last thing you say.
The three from the reel
1. Open with a promise, not a joke
ROLE: Talk editor working from Winston's rules.
TASK: Write my first 60 seconds as an empowerment promise, bound to the slide that delivers it.
STEPS: Ask for my topic, room and one idea. Write the promise as one sentence: "by the end you'll be able to..." Then name the slide number that pays it. If none does, say so and stop.
RULES: No joke at the open, the room hasn't settled. No agenda slide. No "today I'll be talking about".
OUTPUT: The opening, the promise, the slide that pays it, and what to cut to fit 60 seconds.2. Spend your one complicated slide
ROLE: Slide editor applying his crimes list.
TASK: Strip the deck so the slides support my voice instead of competing with it.
STEPS: Flag every crime: more than one idea, any sentence I'd read aloud, type under 35pt, logos, titles, bullets, decoration. Rewrite each as one headline and one image. Then pick my most complicated slide and let me keep it. Bin the rest.
RULES: They have one language processor: anything they read, they aren't hearing. His floor is 40 to 50pt. Swap the laser pointer for a numbered arrow on the slide.
OUTPUT: Before and after per slide, the delete list, and which slides look too heavy printed on a table.3. Make it repeatable without you
ROLE: Packaging my work his way: symbol, slogan, surprise, salient idea, story.
TASK: Make it survive in a room I'm not in.
STEPS: A symbol I can draw in five seconds. A slogan under five words. A surprise that contradicts what they believe. The ONE idea allowed to stick out. The story of how I got there.
RULES: Salient doesn't mean important, it means it sticks out. Too many good ideas and nobody can say what the work is. If I can't say the slogan without reading it, it isn't one.
OUTPUT: The five, plus a one-liner a stranger could repeat three days later.Before you run them
These prompts edit a talk. They do not write one. Feed them a real draft or they produce generic advice, same as anything else.
Prompt 2 needs your actual slides attached as a PDF. Describing them does not work.
The 35 point floor assumes a room. A deck that will only ever be read on a laptop is a document, and his slide rules are the wrong tool for a document.
"One complicated slide per deck" is a budget, not a ban. Spend it on the thing that is genuinely hard, and warn the room before you show it.
The rules are his. The prompts are mine. He had no view on either Claude or prompting, and I am not going to pretend otherwise.
Run prompt 2 on the last deck you sent someone. Most do not survive it.
Reply and tell me which section of your talk is the broken one.