⏱️ Teacher's Workbook: Can AI Actually Make You Faster?

Meetup talk · ~30 minutes · 13 slides · pairs with the projector deck

This is a talk, not a class — no exercises, no timers, no worksheets on every seat. The room is different too: AI meetups and chambers of commerce, which means more skeptics, more people who read the studies, and at least one person who will ask about effect sizes. Good. This talk is built to survive that room, because its whole spine is ranking the evidence by how much you should trust it and admitting the holes out loud. One note on timing: this deck has no clock badge in the corner — the minute marks below are suggested pacing for a 30-minute slot, not a run sheet the deck enforces. Keep your own watch visible, and use the same marks here and in the run sheet so you always know if you're ahead or behind.

Open the projector deck → ← All classes

THE SHAPE OF ITWhat this talk does

People walk in having heard "AI will 10x your life" from someone selling a course. They walk out with three real numbers, one honest hole, and a way to test it on themselves. By the end, the room should be able to:

What this talk deliberately does not do: promise outcomes. AI does not cure, diagnose, or treat ADHD, and no study shows any AI tool improves ADHD outcomes. In this room, the honesty isn't just ethics — it's strategy. A technical crowd has watched a hundred hype talks. Being the one speaker who concedes the limits before they're raised is how you win the room.

BEFORE YOU SPEAKMaterials & setup

Bring:

Logistics: ask the organizer two questions in advance — how long is the slot really (a "30-minute talk" at a meetup is often 25 plus questions), and does Q&A happen in-slot or after. Time-box Q&A out loud: "We've got five minutes for questions, and I'll stay after for the rest." A technical crowd will happily eat twenty minutes on the first question; the time-box is how you serve the whole room instead of the loudest chair. If you're offered a mic, take it — never trust a meetup room's acoustics.

If the slot shrinks: the boulder demo compresses well — two taps instead of four says the same thing. The share-and-chat energy of a class isn't here to cut, so cut warm-up, not content. What never gets cut: slide 5 (the honest hole) and slide 12 (the rules). The honesty slides are the brand.

AT A GLANCEThe run sheet (suggested pacing)

ClockSegmentWhat's happening
0:00TitleThe promise: smaller numbers, better sources, ranked by trust
0:02The best evidenceMIT controlled trial — the gold standard number
0:04Beginners gain most5,179 support agents — the encouraging split
0:06The catchActivity vs. shipped work — the anti-hype number
0:09The honest holeZero rigorous ADHD-specific studies — said plainly
0:11The opportunityGains cluster at starting; ADHD stalls at starting
0:13Where the time comes fromThe four mechanisms — YAP, MAP, TAP, SNAP
0:16The boulderSMALLER demo, on click — the talk's one theater moment
0:18The trapRabbit-holing + the brake prompt
0:20Boring daily winsWhat credible use actually looks like
0:22Become your own studyThree numbers, one week, on paper
0:24The rulesSupports-not-treats, can be wrong, postcard rule
0:26CloseStart 40% sooner — tonight. Done by 0:28, Q&A after

SLIDE BY SLIDEThe segment guide

0:00 Can AI Actually Make You Faster?

Goal: earn a skeptical room's attention in ninety seconds by promising less than every other AI talk they've seen.

Say

"Quick show of hands — who's seen a video promising AI will 10x your life? Keep it up if you believed it. Right. Here's my pitch, and it's the opposite of that video: the real numbers are smaller than the internet says, they come from better sources than the internet uses, and they're still worth having. I spent 29 years behind a bar, which is excellent training for one thing: telling the difference between what a person says and what's actually true. So tonight we're going to do something unusual for an AI talk — we're going to rank the evidence by how much you should trust it, top shelf to well liquor. And where the evidence runs out, I'll say so out loud, because one of those holes is the reason I'm on this stage."

Do

Deliver the whole opening standing still. At a meetup, calm reads as confidence — the hype guys pace.

Watch for

The arms-crossed 10x guy in the second row. Don't spar yet — slide 4 is built for him, and it lands harder as a number than as a comeback.

0:02 The best evidence we have

Goal: plant the gold-standard number — and its provenance — so everything after has an anchor.

Say

"Top shelf first. MIT economists ran a real controlled trial — not a survey, not vibes — with 453 professionals doing writing tasks. With ChatGPT, they finished 40% faster, and here's the part people skip: quality didn't drop. It went up 18%. Published in Science, 2023, and it's still the gold standard study in this space. Now let me put the asterisk on my own slide before anyone else does: it's one trial, on writing tasks, under study conditions. That's what makes it trustworthy and that's also what limits it. Forty percent faster at defined writing tasks is not forty percent faster at your job, and anyone who tells you otherwise is rounding up for the sale. But as a floor for 'does this do anything real?' — this is the number."

Do

Lines reveal one per press — land "40% faster," pause, then hit the quality line. The pause is where skeptics expect the catch; giving them "quality went up" instead is the moment.

Watch for

The skeptic who says "that's lab conditions, controlled tasks — real work isn't like that." Concede it, fully and cheerfully: "You're right, and that's exactly why I ranked it instead of worshipping it." Conceding honest points is the brand; defending past your evidence is how you lose this room.

0:04 Who gains the most? Beginners.

Goal: reframe "I'm behind on AI" from shame into the best seat in the house.

Say

"Second shelf: a field study — real workplace, real stakes — of 5,179 customer-support agents. With AI help, productivity rose 14% on average. Fine number, not a headline. But split it open and it gets interesting: the newest workers gained 34%. The experienced folks gained much less — the AI was basically handing everyone the veteran's playbook, so the people furthest from veteran got the biggest boost. Here's why I love this number for this room and every room: if you feel behind on AI — and half of everyone I meet quietly does — the data says you're not the person this leaves behind. You're the person with the most to gain. The gap between you and the early adopters isn't a moat. It's the discount."

Do

Look at the back half of the room on "if you feel behind" — that's where those people sit at a meetup.

Watch for

An experienced developer hearing "beginners gain most" as "it won't help me." Head it off: the split is about closing skill gaps on defined tasks, not a ceiling on experts — and slide 4 is about experts' numbers next.

0:06 The catch nobody quotes

Goal: detonate the 10x claim with the software numbers — the anti-hype slide that buys you the room's trust.

Say

"Now the number the course-sellers never put on a slide. In software, AI made raw activity jump 40 to 180 percent — commits, code written, motion. But finished, shipped work rose about 30 percent. Sit with that gap for a second, because it's the most important gap in this talk: faster typing is not faster finishing. A lot of what AI accelerates is the middle of work — producing more stuff — and the stuff still has to be checked, integrated, and actually done. So when someone promises you 10x, ask them one question: ten times more activity, or ten times more shipped? Because the best evidence we have says even the activity number tops out under 3x, and the shipping number is 1.3. Real. Useful. Not magic."

Do

Slow down on "faster typing is not faster finishing" — it's the line people quote to a coworker tomorrow. Let it sit.

Watch for

The 10x bro pushing back with an anecdote about his own workflow. Don't argue his experience — "I believe your week. I can't put your week on a slide. These are the numbers that survive an audit, and I'd rather under-promise you." The room sides with the guy holding sources.

0:09 And the honest hole in all of it

Goal: the slide the whole talk is built around. Admit the ADHD research gap plainly, before anyone can ask.

Say

"Here's the slide that makes this a different kind of AI talk. Everything I just showed you was studied on general populations. Not one rigorous study measures AI speed gains for ADHD adults specifically. Not one. So every 'ADHD-ers are 60% faster with AI' claim you've seen on the internet is one of two things: a general-population number wearing an ADHD costume, or a testimonial. I run a site about ADHD and AI, so understand what I'm doing right now — I'm telling you the central claim of my own niche is unproven. I'd rather say that out loud than have one of you find the hole after you trusted me. What we can do honestly is look at where the general gains cluster, and ask whether that location matters for ADHD brains. That's the next slide."

Do

Flattest, slowest delivery in the talk. No jokes on this slide. The contrast with the warmth everywhere else is what makes it credible.

Watch for

The person who wants ADHD-specific numbers anyway — "surely there's something?" Hold the line: there are testimonials and small surveys, and you'll happily call them what they are, but rigorous speed-gain studies for ADHD adults don't exist yet. Do not soften into "well, some early research suggests…" — inventing a study to comfort a questioner is the one unforgivable move in this talk.

0:11 So what do ADHD brains actually get?

Goal: the pivot — locate the real opportunity at "starting" without over-claiming an inch past the evidence.

Say

"So if the ADHD-specific research is a hole, why am I here? Because of where the proven general gains cluster. Look back at every number tonight: writing tasks, first drafts, blank pages, task beginnings. The gains live at the start of work. Now ask anyone with ADHD — I'll save you the survey, I talk to these folks every week — where does the day actually die? Not in the middle of a task. At the start. The staring-at-it, circling-it, three-hours-of-not-beginning part. So here's my careful claim, and notice how careful it is: AI's proven strength and ADHD's biggest bottleneck happen to be the same spot on the map. That overlap is an opportunity, not a study result. But it's a very good place to run an experiment — and the experiment costs nothing."

Do

Hands make this one work: left hand "where the gains are," right hand "where ADHD stalls," bring them together on "same spot on the map."

Watch for

Heads nodding too hard — a friendly room will upgrade your "overlap" into "proof" when they retell it. Repeat the guardrail once more than feels necessary: "opportunity, not outcome."

0:13 Where the time really comes from

Goal: make the mechanism concrete — four specific places minutes come back, mapped to the four words.

Say

"If AI saves an ADHD brain time, here's where the minutes actually come from. One: capturing instantly — the thought goes in the second you have it, by voice, instead of dying on the way to a notebook. We call that YAP. Two: not spiraling — the mess comes back as five piles instead of a storm you re-argue with all afternoon. MAP. Three: starting sooner — a ten-minute action instead of a mountain. TAP. Four — and this is the sleeper — not restarting from zero. Every interruption normally costs an ADHD brain the whole re-load; a save file means you resume instead of restart. SNAP. Notice none of these is 'the AI does your work.' All four are the AI removing the friction around the work. That's the honest mechanism."

Do

Four lines, four presses, one finger per point. Brisk — this is a mechanism slide, not a method class; the meetup crowd wants the model, not the workshop.

Watch for

Someone asking "is this your product?" Plain answer: it's a free method on a free website, there's nothing to buy tonight, and you'll gladly show anyone after. Then move — a sales-pitch smell at minute 13 undoes three honesty slides.

0:16 Watch the mountain shrink

Goal: the talk's one piece of theater — the boulder shrinks on click while you teach the single most reusable word.

Say

"One live thing, thirty seconds, no wifi required. When an AI hands you a plan, it hands you a nine-step wall of text — which, to an ADHD brain, is just a new mountain. There's one word the AI has to obey: SMALLER. Watch the boulder." [Tap — it shrinks.] "SMALLER." [Tap.] "Still too big? SMALLER." [Tap.] "You say it until the task fits in ten minutes, and there's no limit and no shame — five times is fine. I'm only half joking when I say this one word is worth more than most productivity apps, because the apps organize the mountain and this word shrinks it. And it works in any chat AI ever made, because it's not a feature. It's an instruction."

Do

The boulder shrinks on click — you rehearsed the rhythm, so tap-say-pause, three or four beats. If you're running long, this is the slide that compresses: two taps say the same thing as four.

Watch for

Technical folks smirking at the simplicity. Invite it: "Yes, it's a one-word prompt strategy. The people I teach don't need a prompt library — they need the mountain smaller. Simple that gets used beats clever that doesn't."

0:18 The trap that eats every gain

Goal: the rabbit-hole warning plus the brake prompt — the other half of the honesty spine.

Say

"Now the failure mode, because every real tool has one. The same properties that make AI feel wonderful to a stimulation-hungry brain — endless, novel, always answering — make it a world-class rabbit hole. Rabbit-holing with the AI itself: three hours of prompting about the task is three hours the task didn't get, and it felt like work the entire time. That's how a 40% gain becomes a net loss, and I promise you the people in this room with ADHD just felt seen. So here's the brake, a prompt you paste when you catch yourself spiraling: stop giving me new ideas, pick one action from what we have, and don't reply until I say DONE. You're using the AI to end the conversation with the AI. It's on the handout, word for word."

Do

Read the brake prompt slowly, verbatim, and point at the handout stack by the door. This slide never gets cut for time — it's half the reason the talk is trustworthy.

Watch for

Laughter of recognition on "felt like work the entire time" — ride it, don't talk over it. That laugh is the room admitting the trap is real, which makes the brake land.

0:20 Small boring daily wins

Goal: replace the 10x fantasy with the credible picture — unglamorous daily uses that compound.

Say

"So what does credible use actually look like, day to day? Real ADHD coaches — the working ones, not the thumbnail ones — use AI for things like breakfast day-planning, softening a blunt email before it costs a relationship, and beating blank pages. That's it. That's the list. Nobody credible is living a 10x life, and after tonight you know why: the shipping number was 1.3, not 10. What they're doing is starting 40% sooner, over and over, on small boring things — and that compounds, because the cost of not-starting compounds too. I'd rather send you home planning to win breakfast than planning to revolutionize your career by Friday. Breakfast you'll actually do. And Monday's breakfast makes Tuesday's more likely."

Do

Lean into "boring" with a grin — this slide is deliberately anticlimactic, and naming that is the charm.

Watch for

The clock. You're at the two-thirds mark; if you're past 0:23 here, tighten this slide and the close — never the two honesty slides (5 and 12).

0:22 Become your own study

Goal: the payoff of the honest hole — since the ADHD research doesn't exist, hand the room the n-of-1 protocol.

Say

"Remember the hole — no rigorous ADHD-specific study exists. Here's the honest response to that: become your own. One week, three numbers, on paper — paper survives app-switching, which, trust me, is a load-bearing feature for this audience. Number one: taps done — tiny actions actually finished, not planned. Number two: restarts — times a save card got you back into something after an interruption instead of starting over. Number three: time-to-start — minutes between deciding to do a thing and actually beginning. That last one is the money number; it's where the whole opportunity lives. Track a normal week. If the numbers move, you have evidence about the one brain that matters — yours. If they don't, you just saved yourself a subscription. Either result beats my claims."

Do

Hold up the printed handout — "this exact tracker is by the door." Making the take-home physical is what converts a talk into a week of behavior.

Watch for

A data person objecting that n-of-1 without a control isn't rigorous. Agree instantly: "Correct — it's not publishable, it's actionable. You're deciding whether to spend twenty minutes a day, not writing a paper." That concession usually wins the biggest skeptic in the room.

0:24 The rules, one more time

Goal: the guardrail bookend — four rules, stated plainly, so the honesty is the last argument they hear.

Say

"Before the close, the rules — the same ones I open and end every ADHD room with, and this room gets them too. One: AI supports, it doesn't treat. It does not cure, diagnose, or treat ADHD, and no study shows any AI tool improves ADHD outcomes — if a vendor tells you otherwise, they're selling something. Two: it can be wrong. Confidently, fluently wrong, so your judgment stays in the loop. Three, the postcard rule: nothing goes into an AI you wouldn't write on the back of a postcard — fake names, no medical, legal, or money details. Four, the rule that filters everything else: if it doesn't make your next step smaller, it's not helping. Any tool, any model, any year. That rule doesn't expire."

Do

Same flat, serious register as slide 5 — the two honesty slides should sound like each other. Never cut this slide; if you're overtime, shorten the close instead.

Watch for

Chamber-of-commerce rooms sometimes hear the postcard rule as anti-AI. Frame it as professional hygiene: "It's the same rule your lawyer has about email. Using a tool well includes knowing what not to feed it."

0:26 Start 40% sooner. Tonight.

Goal: land the close on the one proven-adjacent number and the smallest possible ask, then hand off to Q&A cleanly.

Say

"Here's everything in one breath. The best evidence says roughly 40% faster on the kind of task where you're staring at a blank page — and blank pages are exactly where ADHD brains lose whole afternoons. The ADHD-specific proof doesn't exist yet, so run the one-week experiment and get your own. The ask tonight is deliberately tiny: one yap, one map, one tiny tap. Talk your mess into a free AI app, let it sort the piles, do one ten-minute thing off the top. That's the entire homework. Everything from tonight — the numbers, the tracker, the brake prompt — is free at adhdyap.com, no account needed. Thank you. We've got about five minutes for questions here, and I'll stick around after for the rest."

Do

The four words land one per press and the slide celebrates on the last one — let it, even in a suit-and-lanyard room. Then restate the Q&A time-box before the first hand goes up; it's much harder to impose after.

Watch for

The organizer signaling time. Land the close anyway — 60 seconds — and move Q&A to the hallway. A finished talk with hallway questions beats a talk that trails off mid-slide.

BE READYTough questions from the room

This crowd asks harder questions than a library class — stats questions, sourcing questions, replication questions. The rule for all of them: concede what's true instantly, never invent a citation to escape a corner, and treat every hard question as free credibility.

"Which study is the MIT one, exactly? What was the design?"

"A controlled trial by MIT economists, published in Science in 2023 — 453 professionals doing writing tasks, randomized to work with or without ChatGPT. The with-AI group finished 40% faster with quality rated 18% higher. I keep the citation on the site rather than reciting author lists from a stage, because misremembering a name in front of this crowd is worse than pointing you to the source. Everything I quoted tonight is exactly as the papers report it — that's a hill I'm happy to be checked on."

"That's lab conditions and assigned tasks. Does it generalize to real work?"

"Partially — and that's the honest ceiling on it. Controlled conditions are why the 40% is trustworthy and also why it doesn't transfer one-to-one to your Tuesday. That's exactly why I showed you the field data next: 5,179 real support agents at 14% average, and software's gap between raw activity and shipped work. Notice the pattern — the closer you get to messy real work, the smaller and lumpier the number gets. That pattern is the takeaway, not any single figure."

"Has any of this replicated? One study isn't evidence."

"Agreed on the principle — one study is a data point, not a law. What I presented is the trust ranking, not a settled literature: one gold-standard controlled trial, one large field study, and messier activity-versus-output numbers from software, each measuring different things on different populations. That's converging evidence that something real happens around task-starting, and it is nowhere near enough to promise you a percentage. Which is why the talk ends with 'become your own study' instead of a guarantee."

"Doesn't checking the AI's output eat the speed gain? It's wrong constantly."

"Sometimes, yes — and the software numbers are partly that story: activity up 40 to 180 percent, shipped work only about 30, and verification tax is part of the gap. Two honest answers. First, the MIT trial measured completed, quality-rated work — the 40% survived the checking. Second, the uses I teach are low-stakes by design: sorting your own brain dump into piles has a near-zero cost of error, softening your own email keeps you as the editor. The higher the stakes of a wrong answer, the smaller your real gain — so we aim the tool where wrong is cheap."

"What about coding agents? My feed says 10x engineers are real now."

"The best numbers I trust for software: raw activity up 40 to 180 percent, finished shipped work up about 30. So the feeling of 10x is real — the motion genuinely explodes — but shipping is the number that pays, and it's 1.3, not 10. Agents are moving fast and tonight's numbers will age; the discipline I'd sell you is permanent, though: whenever someone quotes a gain, ask whether it measures activity or shipped outcomes. That one question deflates most of the feed."

"So there's really not one ADHD study? Not even a small one?"

"Not one rigorous study measuring AI speed gains for ADHD adults specifically — that's the precise claim, and I stand behind it. There are testimonials, surveys, and coach anecdotes, and I use some as illustrations while labeling them exactly that. What doesn't exist is the controlled trial where ADHD adults with AI get compared to ADHD adults without. If you know of one, genuinely, come find me after — I run a site on this and I'd rather correct the talk than protect it. And no study shows any AI tool improves ADHD outcomes, period — which is why every claim tonight was 'supports,' never 'treats.'"

"The support study says experienced workers barely gained. Why bother if I'm already good?"

"Fair reading of the split — 34% for the newest workers, 14% average, so veterans gained least. That study measured a defined skill the AI could hand to novices. But the ADHD case isn't about skill transfer at all — it's about starting friction, which doesn't care how senior you are. A twenty-year expert with ADHD still loses the afternoon to not-beginning. The one-week tracker settles it for your case either way: if time-to-start doesn't move, you have my blessing to skip the whole thing."

"What are you selling?"

"Tonight, nothing — there's no price on anything I showed you. The method, the prompts, the tracker, and these numbers live free at adhdyap.com, no account. Full disclosure since you asked directly: the site has a free ebook that asks for an email, and I teach library classes. That's the whole empire. The honest-numbers talk is not a funnel trick — it's the actual product, because the ADHD audience has been burned by every 10x pitch on the internet and trust is the only thing that works on us."

PRINT MEHandout — become your own study

Print a stack for the table by the door (this section prints clean on its own — use the print button up top). One page: the numbers ranked, the one rule, the brake, and the one-week tracker.

⏱️ Can AI actually make you faster? — the honest one-pager

The evidence, ranked by trust

1. Controlled trial (MIT, Science 2023): 453 professionals, writing tasks — 40% faster, quality up 18%.

2. Field study: 5,179 support agents — 14% average productivity gain, 34% for the newest workers.

3. Software: raw activity up 40–180%, but finished, shipped work only about +30%. Faster typing ≠ faster finishing.

4. ADHD-specific: not one rigorous study measures AI speed gains for ADHD adults. Every ADHD-percent claim you've seen is borrowed or a testimonial.

The one rule

If AI doesn't make your next step smaller, it's not helping. Too big? Say SMALLER — as many times as it takes.

🐇 The brake prompt (copy verbatim when you're rabbit-holing)

"Stop giving me new ideas, pick one action from what we have, and don't reply until I say DONE."

📊 My one-week self-study (three numbers, on paper)

Taps done = tiny actions actually finished · Restarts = times a save card got me back in · Time-to-start = minutes between deciding and beginning

DayTaps doneRestartsTime-to-start
Day 1
Day 2
Day 3
Day 4
Day 5
Day 6
Day 7

If the numbers move, you have evidence for the one brain that matters. If they don't, you saved yourself a subscription.

📮 The postcard rule

Nothing goes into an AI that I wouldn't write on a postcard. Fake names only — Person A, Company B, Project X. No medical, legal, or money details. And: AI can be wrong. Helper, not doctor.

The full method, the prompts, and these numbers, free: adhdyap.com · AI supports an ADHD brain. It does not cure, diagnose, or treat ADHD — and it can be wrong.

AFTERWARDAfter the talk