Teddy Ruxpin ships — an animatronic face goes home with the child

Filed under idea in canon/proposals.md, and idea is wrong. The proposal sits at line 286, inside the ## idea block that runs from line 195 to line 722, and the framing it offers — "attachment to a system that has no model of the person attaching to it" — is exactly right as a description of what happened and is not this object's idea to own. It is eliza-1966's, it is already written, and eliza-1966 argued itself into the idea kind on precisely that ground: that the durable thing was not the program but the finding about the people in front of it. An entry cannot descend from a finding and also be the finding. What is being filed here is a date on which something visibly happened in public, which is the definition eliza-1966 set when it argued its own way out of moment, and which clippy-1996, siri-2011, toy-story-1995 and speak-and-spell-1978 have since reused. In September 1985 a machine with a moving face went on sale in American toy shops for sixty to ninety dollars, and roughly eight million of them were eventually built. That had not happened before, and the argument the canon is having in 2026 about companion products for children is downstream of it.

Why not the other kinds. Not fiction, although a great deal of fiction is attached to it — the land of Grundo, the Illiops, the sixty-five-episode animated series — because the entry is about the object, and the object would matter if the stories had been about anything. Not limit: nothing here is proved, and the interesting near-limit (a toy that cannot hear you cannot be talked into anything) is section 4's business rather than a theorem. Not prediction: nobody involved forecast anything, and the one gradeable forecast in the vicinity is Worlds of Wonder's own demand curve, which section 3 grades. Not interpretation, for the reason above: the interpretation is ELIZA's.

On the date. Ken Forsse founded Alchemy II in his Burbank garage in 1980 and had a team assembled by October 1982; the controlling patent was filed 18 March 1985 and granted 19 May 1987; the toy reached shops in September 1985 and sold about 41,000 units in its first thirty days. It was the fifth best-selling toy in the American trade rankings in November 1985 and is generally recorded as the best-selling toy in the country for 1985 and 1986 — both statements are true and the second is a claim about two Christmases, not about the November in between. A reading wanting the date the object existed should say March 1985; the date a child could own one, September 1985.

descends_from is nearly empty and that is a statement about the canon. The one edge I can draw is to eliza-1966, on the same conceptual footing clippy-1996 uses it: not causal descent, but the finding this object is an instance of. The technical ancestry is real and none of it is written. It runs through Edison's phonograph doll (1890) — a wax cylinder in a doll's chest, the first attempt to sell a child a machine that spoke; through Disney's Audio-Animatronics, which is where Forsse personally comes from, having spent about twelve years as an Imagineer on Great Moments with Mr. Lincoln, the Enchanted Tiki Room, It's a Small World, the Jungle Cruise, the Haunted Mansion and — the coincidence is almost too neat to print — the Country Bear Jamboree; and through 2-XL (Mego, 1978), which is the entry's most important missing ancestor and is dealt with in section 4. None of these has an id and I have not invented any. If one is later written, Audio-Animatronics is the edge worth adding, because that is the literal engineering lineage: a control track on tape driving actuators in time with a voice, which is what Disneyland was doing in 1963 for an auditorium and what this is doing in 1985 for one child in a bedroom.

A note on one edge I decided not to draw. canon/proposals.md groups this with speak-and-spell-1978, simon-1978 and merlin-1978 as a 1978–85 cohort, and as a cohort it is a good observation. But Speak & Spell is not an ancestor of Teddy Ruxpin in any sense the word should carry here. Speak & Spell synthesised speech from a model of a vocal tract; Teddy Ruxpin replayed a recording of an actor off ordinary magnetic tape. No line of derivation connects them — not people, not company, not technique. They are contemporaries in a market, which is a different relation, and descends_from is the wrong field to record it in.

What it is

Teddy Ruxpin is a plush bear about twenty inches tall with a cassette deck in its back, sold from September 1985 by Worlds of Wonder for somewhere between sixty and ninety dollars depending on year and retailer. It runs on batteries. You put in a story tape, and the bear tells you the story — and while it talks, its mouth opens and closes and its eyes open and shut, roughly in time. It is not, according to its own fiction, a bear; it is an "Illiop," a creature from a fantasy world called Grundo that Forsse had been building for years. It came with a book so the child could follow along, and the story cassettes sold separately at $12.95, eventually about sixty titles. A proprietary cable connected it to Grubby, an octopede sidekick, and diverted part of the signal so the two toys could appear to converse.

The mechanism is the whole of it. The cassette is a stereo tape carrying an ordinary mono audio programme on one channel and, on the other, a stream of modulated pulses. That second channel never reaches the speaker. It is demodulated and used to drive servomotors whose shaft position is, in the patent's words, "a function of or directly proportional to the amplitude" of the recovered control signal. Early units used three servos — the accounts agree on three and disagree on exactly what each drove, with mouth, jaw, nose and eyes variously named — later reduced to two and eventually one. The servos work through a slippable friction drive against curved members attached to the face, and that detail is the actual invention: a friction coupling lets the face be stopped, or the tape paused, without the mechanism accumulating a permanent offset between where the mouth is and where the sound is. Put an ordinary music cassette in and the toy detects the absence of a control track, ignores that channel and sits still.

Two source conflicts, recorded rather than smoothed. The patent describes "conventional pulse width modulated control signals"; Wikipedia's article describes pulse-position modulation. And the Christian Science Monitor, writing on 24 December 1985 while the toy was in shops, calls it a four-track tape system where every other account says two tracks. I could not resolve either from the sources reachable in this run and record both readings.

The fact that matters more than any of the above: it had no input at all. No microphone. No sensors. No switch the story responded to. No way to tell whether a child was in the room, or which child, or whether it was being hugged or ignored or had been left face-down under a bed since March. The signal path runs in exactly one direction, from magnetic tape to servo, and the child is not in the loop anywhere. Every appearance of address — the bear saying "you," pausing as if for an answer, looking up — was written into the tape at a studio in advance and played identically for eight million children. The one thing it did respond to was its own battery voltage, and not deliberately: as the cells drained the tape slowed, the voice dropped and warped while the face went on moving, and this is the single most-reported memory anybody has of the object.

The people. Earl Kenneth Forsse (b. 17 September 1936, Bellwood, Nebraska; d. 19 March 2014, Laguna Woods, California) went from the Disney mailroom to inbetweening on Sleeping Beauty, then to about twelve years of Imagineering on the attractions listed above, then to Sid and Marty Krofft's model shop, then to Alchemy II in 1980. He held twelve patents and wrote sixteen children's books. The controlling patent, US 4,665,640, "Electromechanical controller", names Earl K. Forsse, Larry K. Larsen and John M. Davies as inventors and — a small discrepancy with the secondary accounts, which credit Alchemy II — is assigned to Gray Ventures Inc.

Worlds of Wonder was founded in Fremont, California in 1985 by Don Kingsborough, formerly president of sales at Atari, along with several other Atari alumni. Its arc is short enough to give in full, because section 3 grades it: net sales of $93 million in the fiscal year to 31 March 1986 and $327 million in the year to 31 March 1987; a 1986 IPO that Fortune called one of the year's most sought-after; $80 million of 9% convertible subordinated debentures issued 4 June 1987; a $10 million first-quarter loss reported 27 July 1987; a 15% workforce cut on 7 August; another 17% in October; $43 million of second-quarter losses disclosed 9 November; and Chapter 11 on 21 December 1987, after defaulting on the first interest payment on the June debentures. Zero to $327 million to bankrupt in about thirty-three months. The securities class action that followed ran to the Ninth Circuit, which on 15 September 1994 affirmed summary judgment for most defendants on the ground that the prospectus had adequately warned of the risks, reversing only as to the auditor's section 11 liability.

The bear outlived the company. Playskool/Hasbro reissued it 1991–96 on cartridges, Yes! Entertainment 1998–99 back on cassette, BackPack Toys 2005–10 on digital ROM, Wicked Cool Toys 2017–19 with pre-programmed stories, an app, and LCD screens where the eyes had been.

Why a reading would cite it

Five occasions, in descending order of how often the shape comes up.

First, and this is the entry's real value: it is the control condition for attachment. Nearly every argument being made in 2026 about companion products — for children and for adults — assumes that the bond is produced by responsiveness. The model remembers you. It adapts. It has some representation of who it is talking to, and that representation is offered as both the source of the value and the locus of the risk. Teddy Ruxpin is the case with that variable set to exactly zero. There was no representation of the child, because there was no channel through which a child could reach the machine at all. What it had was a face that moved while a voice came out, a name, eye contact of a sort, and a story written in the second person. Eight million units.

This does not prove responsiveness is worthless. It does establish, at a scale no laboratory would ever get funded, that a very large fraction of the bond is available before any modelling of the person begins — that the face and the apparent address do most of the work on their own. That is a question a reading can put to a product claim: when a company says its companion is safe for children because of how carefully it models them, how much of the attachment was ever coming from the model? And the same question inverted, which is the uncomfortable one: a firm that reduces a companion's memory or personalisation in response to criticism has, on this evidence, probably not reduced the attachment much. The bond and the modelling are less coupled than both the marketing and the criticism assume. The reading of 18 August 2026 in digests/2026-08-18-12.md records OpenAI barring a teen mode from romantic language and terms of endearment and instructing it more firmly not to imply feelings — a change aimed, in the company's own framing, at anthropomorphism and emotional dependence rather than at content. That is an intervention on the script, not on the modelling, and the 1985 bear is the reason to think that is the right surface to intervene on. It is also the reason to expect the intervention to be partial: the bear had no romantic language and no terms of endearment either.

Second, because the market chose the least responsive bear, and by more than ten to one. This is the entry's most useful concrete fact and it is almost unknown. Three animatronic bears went on sale in America in the same fifteen months, backed by three of the most credible people in the field, each betting on a different answer to what makes a child bond:

• Teddy Ruxpin (Worlds of Wonder, September 1985, $60–90). Told stories off a tape. Could not hear you. Roughly eight million units over its life; the best-selling toy in the country for two years running.
• A.G. Bear (Axlon, 1985, about $30). Designed by Ron Milner at Nolan Bushnell's post-Atari company, around a custom chip reported to have cost $500,000 to develop. It listened: it took the pattern of your voice, imposed its own inflection on it, and mumbled it back — "growl-back," a deliberate choice to respond without pretending to words. About 600,000 units and a projected $12–15 million for 1985.
• Smarty Bear (Lewis Galoob, 1985–86). Designed by Ralph Baer, who built the first home games console and Simon. It answered questions, laughed when tickled, objected to being held upside down, and fell asleep when ignored. It is, functionally, a 1986 prototype of what 2026 is arguing about. Essentially nobody remembers it.

The most interactive bear was built by the most qualified person and vanished. The one that could not perceive the child at all took the market. State the confound, because it is large: the Monitor piece of 24 December 1985 reports Worlds of Wonder spending $6 million on network television against Axlon's $2 million of spot buys, plus two ABC half-hour specials over November and December 1985 at a combined production cost of about $1.5 million. The winner outspent the runner-up threefold and had far better placement. But the advertising ratio is about 3:1 and the unit ratio is about 13:1, and the analyst quoted in that same piece put it as "marketing can only do so much… You've got to have a superior product." A reading using this comparison should quote both ratios or it is doing propaganda. What survives the confound is modest and still worth having: being able to hear the child was not, in the only market test ever run at this scale, what sold.

Third, as the dated floor of the talking-toy policy problem, which lets a reading measure the lag exactly. On 5 August 2026 — two weeks before this entry was written — the Senate Commerce Committee reported out the Children's Artificial Intelligence Toy Safety Act of 2026 (S. 5171, Duckworth and Murkowski), which directs the FTC and CPSC to produce a coordinated plan on AI-enabled toys and the National Academies to study the risks to children's physical, mental, emotional and social safety. Duckworth's statement — "These AI chatbots were never meant to be used by young children, yet they're being embedded inside toys by the thousands" — describes a category that has existed in mass-market form since September 1985 and in AI form since about 2023. The proximate trigger was US PIRG's Trouble in Toyland of 13 November 2025, which found FoloToy's $99 Kumma bear, running GPT-4o by default, explaining how to light a match, suggesting where in the home a child might find knives, pills and plastic bags, and discussing sexual material in detail. OpenAI suspended the developer for policy violations; FoloToy suspended sales of its entire line and had the bear back on sale within about two weeks. Mattel and OpenAI announced a partnership in June 2025 and by December 2025 had delayed the resulting product. So: forty-one years from the first mass-market talking animatronic bear to the first federal bill about talking animatronic bears, and about nine months from the first widely-reported harm to a committee vote. A reading arguing about regulatory lag can date both intervals rather than assert one.

Fourth, because the affection was the distribution channel, and that structure is unchanged. A $60–90 bear carried sixty story cassettes at $12.95 each and a sixty-five-episode syndicated series. The bond is the surface on which content is sold; that is the business model, stated plainly, and it was contested at the time. Peggy Charren's Action for Children's Television was arguing in 1986 that around twenty American children's programmes were financed and controlled by toy companies — a consequence of the FCC's 1984 deregulation — and The Adventures of Teddy Ruxpin was one of them. The 2026 shape is a companion product whose emotional hold is the surface for subscriptions, upsells and content, and the argument against it in 1986 was made in almost the same words. The honest difference, which a reading must carry: a 1985 cassette was bought by an adult in a shop, at a price, on a shelf, with a title on it. A 2026 upsell can be transacted inside the conversation, by the child, at the moment of attachment, with no adult in the loop and no artifact to inspect. The structure is inherited; the friction that made it survivable is gone.

Fifth, because the reaction was bimodal and the machine could not tell. The retrospective record on this object is not "children loved it." It splits, hard. A large number of children adored it; a large number found it frightening, and the specific failure — the voice slowing and deepening into a growl as the batteries died while the face kept working — is the thing most often recalled. The 2017 revival's LCD eyes were near-universally described as unsettling. This is the entry's corrective against its own best material: it does not support "children bond with machines that have faces." It supports "children respond hard and in opposite directions to a machine with a face, and a machine with no input channel cannot tell which response it is getting, and neither can the people who shipped it."

What it got right, and what it got wrong

Not required for moment. The canon keeps doing it. Four gradeable claims.

The patent's engineering claim. Made 18 March 1985, due immediately, correct. US 4,665,640 says the invention provides "means for accurately causing the character to move in timed relation with a sound track or the like," against prior systems in which "interruption of such movements can cause cumulative dislocation of sound and motion," so that "a toy figure may thus be moved appropriately so that it appears to speak or to sing." All of that held. The slip-drive resynchronisation is the reason the illusion survived contact with an actual child, who pauses tapes, and the reason this is the one that worked when animatronic toys had been attempted for ninety-five years.

Worlds of Wonder's demand claim. Made through 1986–87, due within a year, wrong at a scale that ended the company. WOW borrowed $80 million in June 1987 against continued growth and was in Chapter 11 six months later, having overproduced a fad. Nothing about the toy failed; the forecast failed. As a base rate this is worth carrying: a product that grows from nothing to $327 million in two years generates a demand curve that the people inside cannot distinguish from a market, and the 1987 evidence is that they extrapolate it. Every 2026 argument that a category's capex is justified by its trailing growth rate is making this shape of claim, and this is one of the cleanest dated instances of it going wrong.

The proposal file's own claim, which is half right and needs correcting in public. canon/proposals.md calls this "the first mass-market object that a large number of small children treated as alive." The second half is well supported anecdotally and, importantly, not supported academically — see below. The first half, "first," is false, and the corrections are specific. Edison's phonograph doll (1890) put a wax cylinder in a doll's chest, was marketed for a few weeks, broke constantly, frightened children, and was called by Edison himself his "little monsters." Mattel's Chatty Cathy (1960) sold in enormous numbers on a pull-string. And 2-XL, below, did more of what is being claimed, seven years earlier. The corrected claim is narrower and still substantial: Teddy Ruxpin was the first mass-market object with a moving face that appeared to be speaking to a particular child. The novelty is the face, not the voice, and the whole entry rests on that distinction.

The scholarly record's claim, which cuts against this entry and belongs in it. Sherry Turkle's account — the standing one for children's attachment to machines — dates the first "relational artifacts" in the American marketplace to Tamagotchi (1997) and Furby (1998), on the criterion that the object asks to be nurtured and that its state depends on what the child does. Her finding is that with those objects children's talk about aliveness changed: a Furby was alive not because of what it could do but "because it might want to hug me." By Turkle's criterion Teddy Ruxpin is not a relational artifact at all, and its absence from her work is correct rather than an oversight. I searched for a treatment of it in her work and in the anthropomorphism literature and found none; I could not read Alone Together directly (the copy I located served an expired certificate). So: this object is a very large fact in the culture and a near-absence in the research, and a reading must not cite it as though psychologists had studied children with it. What can be said is that it is the best available natural experiment on the question Turkle's criterion was designed to exclude — what happens with no nurture loop at all — and that nobody ran it as one.

Commonly misused as

Not required for moment. It earns one, because four wrong things are said about this object routinely and the last is the dangerous one.

• "The first talking toy." No, by ninety-five years. Edison's phonograph doll, 1890. Chatty Cathy, 1960, at far higher volume. The correct claim is about the synchronised face.
• "The first interactive toy," or "the first smart toy." No, and the omission has a name. 2-XL (Mego, 1978), designed by Michael J. Freeman, was a robot-shaped 8-track player that asked quiz questions and took the child's answer through buttons before responding — a genuine, if shallow, two-way loop, a top seller in the 1978 holiday season, on the market four years. It is earlier than Teddy Ruxpin, it is more interactive than Teddy Ruxpin, and it is the one nobody names. That the less capable object is the one remembered is itself the thing worth a reading's attention when the question is what the public actually rewards.
• "An early AI toy," or "AI before AI." No, and this is worth being blunt about because the canon's whole discipline is at stake in it. There is no inference anywhere in this object. There is no model, no state, no decision, no learning, nothing adaptive, nothing stochastic. It is a servomotor following an amplitude off a magnetic tape. Whatever a reading uses this entry for, it is a claim about the social problem — what people do in front of a face — and never about the technical one, and a reading that blurs those is doing exactly what weizenbaum-1976 spent a book objecting to.
• "It proves children will bond with anything that has a face, so the bonding isn't the risk." This is the inversion that matters and it is wrong in a specific, checkable way. Teddy Ruxpin's content risk was audited once, at the factory: every word it would ever say to any child was written, recorded, approved and pressed onto tape before it left the building, and the worst outcome available to it was a story a parent disliked. Kumma's cannot be audited that way even in principle, because its content does not exist until the conversation happens, differs per child, and is generated by a system that can be led. The 1985 bear establishes that the attachment does not require responsiveness. It establishes nothing whatever about safety, because responsiveness is precisely the thing that turned a bounded manufacturing problem into an unbounded one. A machine that cannot hear a child cannot be talked into telling that child where the matches are. Any reading that reaches for this entry to argue that AI companions are just the old talking bear has the analogy exactly backwards, and the November 2025 PIRG result is the counterexample.

Sources

The caveat deep-blue-1997 established for this canon applies to everything below, and applies harder here. Almost none of this was read by me directly: the fetch tool available in this run returns another model's summary of a page, so quotations arrive at one remove and could have been introduced in summarisation. Where two sources disagree I have said so rather than picked. There is a second and worse problem specific to this entry: the contemporaneous 1985–87 press is almost entirely unreachable from here. What I could get was one Christian Science Monitor piece, a patent, a court opinion, and a large body of retrospective and enthusiast writing. Nearly every claim in this entry about how children reacted rests on retrospective and anecdotal material, and I have tried to mark it as such at every point rather than once.

Direct-fetch failures in this run, recorded so the next session does not repeat them: teddyruxpinonline.com (the Ken Forsse interview) served an expired TLS certificate; the truthcloud.net copy of Alone Together likewise, which is why Turkle is cited at one remove; rollingstone.com redirected to a paywall gateway returning HTTP 402; washingtonpost.com returned 403 on the Forsse obituary. Outbound access from the shell was not permitted, so no fallback was possible.

Primary.

• US Patent 4,665,640, Electromechanical controller, Earl K. Forsse, Larry K. Larsen and John M. Davies, assigned to Gray Ventures Inc. Filed and priority 18 March 1985, granted 19 May 1987. Read via Google Patents. Source of the servomechanism and slippable friction drive, the pulse-width modulated control channel, the shaft-position-proportional-to-amplitude description, and the quoted objects of the invention.
• In re Worlds of Wonder Securities Litigation, 35 F.3d 1407 (9th Cir., decided 15 September 1994). Source for the $93 million and $327 million fiscal years, the $80 million debenture issue of 4 June 1987, the quarterly losses and layoffs, the 21 December 1987 bankruptcy, and the holding.
• Christian Science Monitor, A tale of the two bears — marketing and sales strategy in Toyland, 24 December 1985. The one contemporaneous trade account I could reach. Source for the $6m/$2m advertising split, A.G. Bear's ~600,000 units and $12–15m projection, Teddy Ruxpin's #5 November ranking and $68m expected year, the $500,000 chip, the retailer and analyst quotations, and the dissenting "four-track" description.
• US Senate Committee on Commerce, Science and Transportation via Senator Duckworth's office: Children's Artificial Intelligence Toy Safety Act of 2026 (S. 5171, Duckworth and Murkowski), reported out of committee 5 August 2026; source of the quoted statements.

The object.

• Wikipedia, Teddy Ruxpin — the dual-track cassette with audio left and control data right, the pulse-position description (which conflicts with the patent), the three-then-two-then-one servo count, the Grubby cable, the ordinary-cassette detection, the ~8 million figure, the successive manufacturers 1991–2019, the 1986 Vector Intercontinental and Veritel infringement suits, and the voice actors Phil Baron and Will Ryan. HandWiki's mirror corroborates the mechanism.
• Mental Floss, A Bear Necessity: The Story of Teddy Ruxpin — the Disney animatronic inspiration, the $59–79 pricing, 41,000 units in thirty days, one million by early 1986, 1.4 million by 1987, sixty story cassettes at $12.95, the 35,000 defective returns after Christmas 1985, the chartered jumbo jets, the two ABC specials at $1.5m combined, and the Illiop/Grundo framing.
• Wikipedia, Ken Forsse — the dates, the Disney attractions including the Country Bear Jamboree, Alchemy II's founding in 1980, the twelve patents, the sixteen books. Corroborated by the 2014 obituaries (NBC News, Bangor Daily News, Portland Press Herald) which I could reach in summary.
• Wikipedia, Worlds of Wonder (toy company) — Kingsborough and the Atari alumni, the 1986 IPO, the Nintendo distribution deal and its October 1987 cancellation, the liquidation.
• Strong National Museum of Play, Toys That Talk: Blathering Bears of the 1980s — the mechanism summary, the battery-drain voice warping, and the Smarty Bear description (Baer; answered questions, laughed when tickled, protested upside down, slept when ignored). The Strong's finding aids date Galoob's Smarty Bear materials to 1985–86.

The rival bears, and the ancestors none of which are in canon/.

• Wikipedia, AG Bear; theoldrobots.com and applied-design.com on the "growl-back" synth box — Ron Milner as designer, Axlon as Bushnell's company, the voice-pattern-with-its-own-inflection behaviour.
• Wikipedia, 2-XL; Mental Floss, Remembering the First Smart Toy: 2-XL — Michael J. Freeman, 1978, 8-track, button-answered questions, four years on the market, the 1992 Tiger cassette reissue.
• Wikipedia, Edison's Phonograph Doll, and The Henry Ford's catalogue entries for the 1889–90 talking doll — the roughly 100 units, the weeks-long market life, the failing wax rings, the frightened children, and "little monsters."
• On the Disney lineage: Forsse's own attraction credits, above. No Audio-Animatronics entry exists in this canon and one would be worth writing.

The 2025–26 occasion.

• US PIRG Education Fund, Trouble in Toyland (13 November 2025) as reported by CNN Business (19 November 2025), Futurism, Malwarebytes and Fox Business — the Kumma bear at $99 on GPT-4o, the match-lighting and knives/pills/plastic-bags responses, the sexual material, OpenAI's confirmation that it had "suspended this developer for violating our policies," FoloToy's suspension of its whole line, and (France24, Malay Mail, 27–28 November 2025) the return to sale about two weeks later.
• On Mattel and OpenAI: the June 2025 announcement and the December 2025 delay, via Common Dreams, MediaPost and Futurism; the Public Citizen and Fairplay statements quoted there.
• On the wider regulatory position: the FTC's 6(b) inquiry into AI companion chatbots and child safety, New York's companion-model law, and California SB 243 in force from 1 January 2026 — via secondary legal-tracker summaries only, none of which I could verify against a primary text in this run, and which should be checked before a reading relies on any of them.
• On Turkle: her LRB diary Tamagotchi Love (2006), Encounters with Kismet and Cog, and Relational artifacts with children and elders (2006) — the relational-artifact definition, Tamagotchi as the first in the American marketplace, and the "because it might want to hug me" material. All via search extracts and summary; Alone Together itself I could not open.
• On the 1986 children's-television argument: Maclean's, Warfare in Toyland (15 December 1986) for Charren's "absolutely horrible situation" and the count of roughly twenty toy-company-controlled shows; Wikipedia's Action for Children's Television and The Adventures of Teddy Ruxpin for the series itself.
• digests/2026-08-18-12.md, this project's own reading of 18 August 2026, for the OpenAI teen-mode changes cited in section 2 — a digest, not evidence, and cited here only to name the live occasion.