Artificial minds.
Mechanical reasoning, automata, the Mechanical Turk, cybernetics, expert systems, connectionism, the two AI winters. The current moment is a chapter, not a beginning.
Every cycle mistakes fluency for understanding, then discovers the gap. The interesting question is never whether it is intelligent. It is what people did with it while they believed it was.
Joseph Racknitz, Ueber den Schachspieler des Herrn von Kempelen, Leipzig 1789 · public domain
Pick two or three concrete waypoints you can actually talk about rather than listing everything. The Mechanical Turk is worth a minute on its own: a fake automaton with a person hidden inside is a startlingly good description of several products currently shipping.
On the plate: this is Racknitz's cut-away, not a photograph — he was reconstructing how he thought the trick worked, and he got the mechanism wrong. One sentence is worth it: the earliest published explanation of the fraud is itself a guess. Point at the operator in the cabinet, then leave him there — slide 27 is the same diagram, 234 years later.
Artificial realities.
Panoramas, stereoscopes, Sensorama, the Sword of Damocles, VRML, Second Life, Oculus, spatial computing. Roughly a thirty-year cycle of the same promise.
The technology gets better every cycle and the adoption problem never changes, because the adoption problem was never resolution or latency. It is that you have to put something on your head.
Morton Heilig, US Patent 3,050,870 (1962), fig. 5 · public domain
The patent figure does the argument for you. Sixty-four years old, and the thing the drawing is built around is a hood you put your head into. Nothing about resolution or latency is visible in it, because neither was ever the problem.
Sutherland's Sword of Damocles is the funnier image — a ceiling-mounted rig holding the display above the wearer's head — but it is © Harvard and not free. If you want it, it needs a clearance request; this patent drawing is rights-clean and makes the same point.
3D television.
LG Electronics · CC BY 2.0
On the plate: the accessory is the case. Every one of these had to be bought, charged or handed out, worn over prescription glasses, and kept track of for each person on the sofa. Ask how many they would need to own for a family film night. That question is the whole slide.
Exit dates vary by brand and by whether you count "announced" or "last shipped" — Vizio 2014, Samsung 2016, Sony and LG January 2017. Say "by 2017 nobody was shipping one" rather than defending a single date.
Google Glass.
Note what survived: the version nobody had to wear in a bar. Same device, different room, eight more years.
Steve Jurvetson, Google HQ, Jan 2013 · CC BY 2.0
On the plate: two generations in one frame. She is wearing the shipping product; the rig on the screen behind her is where it came from. The gap between those two objects is fifteen years of making it socially wearable — and it still was not enough.
This is the clearest teaching case in the set after the Picturephone. The failure is legible to everyone in the room without any technical knowledge, and the enterprise survival is the twist that makes the point precise: it was not the object, it was where you had to wear it.
Segway.
But the form factor won. The same balance-and-battery stack came back as the shared e-scooter, and the company that killed the Segway now sells those at scale. The technology was never the problem.
Top: Tony Hisgett, Washington DC, 2016 · CC BY 2.0. Bottom: CAPTAIN RAJU, Lisbon, 2020 · CC BY-SA 4.0
On the plate: put the two frames side by side before you say anything. Top: helmets, a guide, single file, a fee, a booking. Bottom: the same balance-and-battery promise, parked on a pavement, unlocked with a phone, nobody supervising and nobody looking ridiculous. Same technology, and the entire delta is social packaging.
The scooter coda is the whole reason to include Segway. Without it this is just a joke about a funny vehicle; with it, it is a case about a correct technology in the wrong social package, ten years early.
NFTs.
ERC-2981, NFT Royalty Standard · CC0
Read the sentence off the plate, out loud, verbatim: “The royalty payment must be voluntary, as transfer mechanisms such as transferFrom() include NFT transfers between wallets, and executing them does not always imply a sale occurred.” That is the standard itself, marked Final, saying the feature was never guaranteed. Not a critic, not a journalist, not me. The primary source is the argument — and it was published in 2020, two years before the money arrived.
Do not put a Bored Ape on this slide. Yuga Labs' own 2023 filings assert common-law copyright while conceding no registration for the images; it is contested, unlicensed, and the cliché. This screenshot is CC0 and a better slide.
Keep the emphasis on royalties, not on prices. The price collapse is the boring part and they already know it. The royalty story is a design lesson: the feature that justified the whole thing was never actually built into the thing. It lived in marketplace policy, so it evaporated under competition. Ask them where else they have seen a promise held up by a norm rather than a mechanism.
Sources disagree on the peak — use annual totals, not headline collapse percentages.
The metaverse office.
This one died six months ago. It is not history — you watched it happen.
Meta Newsroom, Horizon Workrooms launch, Aug 2021 · press use
On the plate: this is Meta's own launch image, not a critic's screenshot — chosen by the company to look its best. Legless cartoon avatars around a conference table. Let the room react before you say anything; the laugh is the argument.
The detail worth naming: Meta published this in August 2021, shut the product down in February 2026, and the launch post is still live on their newsroom. Pull it up beside the discontinuation notice if you have the network.
Nobody published adoption numbers, so do not quote any. The point stands without them.
The recency is the value here. Every other case is something they inherited as received wisdom; this one they lived through, which makes the pattern feel like a tool rather than a story.
The Picturephone, 1970.
Fifty years later the identical capability succeeded, unchanged in function, once it was free and optional.
DogsRNice, Early Television Museum, 2022 · CC BY-SA 4.0
On the plate: a museum object with a loan card next to it. That is where a working technology ends up when nobody wants it — not broken, not superseded, just declined. Worth saying: the device on this table did the thing your phone does, in 1970, and it did it well.
The better image is not free. IEEE's history wiki (ethw.org/Picturephone) holds the 1970 AT&T advertisement showing one actually in use, credited "Courtesy AT&T Archives and History Center" — permission-based, and IEEE is a clean party to ask. Worth a request if you teach this again.
This is the case that isolates the variable. Same function, same promise, opposite outcome. The only things that changed were price and social norm — not capability. If they take one case out of the room, it should be this one.
Thirty-two sets. Let that number sit for a beat before you move on.
Hype cycles are a genre of fiction.
You have seen the curve with the peak and the trough. It is worth knowing that it is a commercial product sold by a research firm, not an observed law. Nothing measured produces that shape, and technologies do not reliably travel along it.
It is not useless. It is a rhetorical device for talking about expectation. It is just not evidence, and it gets cited as if it were.
Jeremy Kemp, volunteer redraw · CC BY-SA 3.0
Worth being precise and fair here rather than dismissive — the curve is a genuinely useful communication tool. The failure is in how it gets used: as a prediction with a position on it, when it has no predictive content.
Say where this particular drawing came from — it is the best beat on the slide. “Hype Cycle” is a Gartner trademark and Gartner's own curves are proprietary, sold. The only reason a free version exists at all is that volunteers redrew it as though it were a natural law. So what is on screen is a fan reconstruction of a consulting deliverable that has been mistaken for physics. The image is an example of the thing the slide is criticising.
Two better tools.
Adoption is slow, then fast, then saturated. Useful because it asks a real question: where on the curve is this, and what evidence tells you? Most things called emerging are still flat, and flat can last thirty years.
We overestimate the effect of a technology in the short run and underestimate it in the long run. Almost everything in the graveyard is the first half. Some of it will come back as the second.
The scooter is Amara's Law with wheels. So, probably, is the headset.
After Rogers, Diffusion of Innovations, 1962 · public domain
The plate is the S-curve only. Amara has no diagram and should not get one — he never published a chart, the “law” is a single sentence, and every diagram in circulation is an uncredited blog redraw. Importing one would hand a random illustration the authority of a citation, which is exactly the failure this lecture is about. If you want one, draw it yourself: two lines, expectation against actual impact, five minutes in vector.
Use the plate to ask the real question rather than to explain the curve: where on this is the thing you are excited about, and what evidence puts it there? Most things called emerging are still flat, and flat can last thirty years.