Simon launches at Studio 54 — a machine that takes turns with you, and nothing inside it that plays

The id is a collision, and it has to be said before anything else. simon-1978 is not Herbert Simon, who won the Nobel Memorial Prize in Economic Sciences in 1978, and whom this canon already holds at simon-1965 — The Shape of Automation. Two Simons whose ids differ by two digits, and the one who is not in this file won a Nobel in the very year the id carries. It is also not Edmund Berkeley's Simon of 1950, which section 4 deals with at length because the coincidence turns out to be worth something. A reading that meets simon-1978 in a descent list and reaches for bounded rationality will be wrong, and nothing in the id warns it. If the owner can still rename a proposal, simon-game-1978 is the safer id; the job names this file, so I have written the one I was given and flagged it here.

Filed under idea in canon/proposals.md, and idea is wrong. The proposal sits at line 301, inside the ## idea block that runs from line 195 to line 722, and what it offers as the idea — "a system with almost no internal complexity is described in mental terms, which is the cheapest and oldest version of the mistake eliza-1966 documents" — names, in its own sentence, the entry that already owns it. eliza-1966 is filed as idea. An entry cannot descend from a framing and also be the framing; that is the argument speak-and-spell-1978 made about the AI effect and dijkstra-1959, and it applies here without modification. This is the fourth consecutive member of the 1978–85 toy cohort to take this correction, after speak-and-spell-1978, teddy-ruxpin-1985 and merlin-1978, and merlin's file guessed that four identical corrections are a fact about the proposal file rather than about the objects. Having now made the fourth, I can say what the fact is: all four are objects, and an object is almost never an idea. An object is a date on which something became purchasable, which is moment by this project's own definition — "something that happened and changed what people expected next," in build.py's words.

Why not the other kinds. Not fiction: there is no story, only a slider and four bulbs. Not prediction: nobody involved forecast anything, and the three gradeable claims in the vicinity — Baer's about his tones, Milton Bradley's about its product, and a television jingle sixteen years late — are section 3's business. Not interpretation. Not limit, and this one deserves a sentence, because there is a hard limit sitting in the middle of this object and it belongs to psychology rather than to Milton Bradley. Simon's top skill level asks a person to reproduce 31 signals in fixed order. George Miller's 1956 paper on the span of immediate memory — not in canon/, and I have invented no id for it — puts unaided recall at around seven chunks, a figure that is routinely over-applied and that certainly does not mean 31 is impossible; it means 31 is only reachable by chunking. The limit is real, it is the player's, it predates the toy by twenty-two years, and what Simon does with it is section 2's second item.

On the date, which is layered. Baer and Morrison saw Atari's arcade Touch Me at the Music Operators of America trade show in 1976 and spec'd their own version, called Follow Me, the following year. The utility patent was filed 19 September 1977 and granted 10 June 1980. Milton Bradley launched the finished product at Studio 54, at midnight on 15 May 1978, with a four-foot Simon lowered over the dance floor. It was short in shops by Christmas 1978. It was Milton Bradley's top-selling item by 1980. Two respectable business histories — encyclopedia.com's and referenceforbusiness's, which are the same text — date the release to 1977, and they are wrong; that is the year of the patent filing, not the launch. Toy Tales prints the filing date as "September 19, 1997," which is a typo for 1977. The id's year is right and the id's month is 15 May.

descends_from, and one edge I have declined. The single edge is to eliza-1966, drawn on the conceptual footing clippy-1996, teddy-ruxpin-1985 and merlin-1978 established, and it is the honest one: Weizenbaum's finding is what happens in front of this object, and the manual quoted in section 1 is Milton Bradley writing it down.

I have not drawn the edge to shannon-chess-1950, which merlin-1978 drew and which would have been the easy consistency. Merlin earned it: it contains a tic-tac-toe player, a blackjack dealer and a Mastermind solver, and those are instances, however cheap, of Shannon's proposition that a machine can be made to play a game against a person. Simon contains no such thing. There is no evaluation, no search, no policy, no representation of the game's state beyond a list, and no representation of the player at all. Drawing the Shannon edge would assert a descent that does not exist and would bury the one fact this entry is for — that the experience of an opponent turns out not to require Shannon's machinery, or any machinery. The refusal is the finding.

The true causal ancestor is Atari's Touch Me (1974), which is not in canon/ and for which I have invented no id. speak-and-spell-1978 and merlin-1978 are contemporaries and not ancestors — same year, same Texas Instruments part family, no line of derivation — and the order matters in one direction: Simon shipped in May 1978 and Merlin in the autumn, which merlin's own file states against itself.

What it is

Simon is a black plastic saucer, ten and a half inches across and two and a half high, that sits on a table between people. The top face is quartered into four translucent coloured lenses — green, red, yellow, blue — each with an incandescent bulb behind it. Not LEDs: bulbs, type 14A, 2.5-volt screw base, and the manual devotes a page to replacing one, noting that "SIMON comes equipped with a spare bulb stored inside the unit." It runs on one 9-volt transistor battery for the electronics and two D cells for the lamps. In the middle is a control panel: an ON/OFF slider, a GAME SELECTOR for games 1, 2 and 3, a SKILL LEVEL slider for 1 to 4, and three buttons — START, LAST and LONGEST. The circuit boards were made in Singapore.

Inside is a Texas Instruments TMS1000, the 4-bit calculator-class microcontroller that was in half the electronic toys of the decade; MAME's dump of the revision-A board carries the mask designation MP3226. Its entire world is 1,024 bytes of mask ROM and 64 four-bit words — 32 bytes — of RAM. The same family, one step down, that merlin-1978 runs on, and the same family that carried Speak & Spell's controller.

The four tones are a bugle. Baer went looking for notes that could not clash, found the answer in his children's Compton's Encyclopedia, and took the harmonics a bugle can sound — which is why, as he put it, they are "pleasing in any sequence" and, in Wikipedia's phrasing, "designed to always be harmonic, no matter the sequence." The shape is a major triad in second inversion plus the octave. The key is not settled by the sources: Wikipedia describes it as an A major triad and names the notes E, C♯, A, E, and records that some 1978 units shipped with a B♭ minor set instead; IEEE Spectrum describes the same shape in C, as G, C, E, G; and the one engineer who put an oscilloscope on a unit measured 209, 252, 310 and 415 Hz — G♯3, B3, D♯4, G♯4 — and observed that the tuning is inaccurate anyway because of "the constraints of accurately producing the duty-cycle timing required for the notes." Those four measured pitches are a G♯ minor triad plus its octave, which is my arithmetic on his numbers and not his claim, and which puts his unit in the minor family rather than the major one. Wikipedia gives no frequencies at all. What is solid is the bugle and the interval shape; anyone quoting Hz should say which unit they measured.

What it does is one thing. From the manual's opening line: "Think fast... SIMON says, 'Chase my flashing lights and sounds'! The challenge is to repeat the ever-increasing random signals that SIMON generates." The machine lights a lens and sounds its note; you press it back; it repeats that signal and adds one; you repeat both; and so on. The skill level sets the length that counts as a win — 8, 14, 20 or 31 signals. "Notice that after the 5th, 9th and 13th signals in a sequence, SIMON automatically increases the tempo." You have five seconds to answer each signal, per the manual; the engineer above measured three on his unit. Fail either test and "SIMON responds with a 'RAZZ' sound; you have lost." Win at levels 1 to 3 and "SIMON will salute you with six short signals from the last lens you played." Win at level 4 and, in the manual's own words, "SIMON blows its mind, loses its memory and gives itself the RAZZ sound. Congratulations, you have beaten SIMON!" Game 2 has the player build the sequence instead; game 3 assigns one lens per player and eliminates them one at a time. In every mode, "Players may team up against SIMON or compete with each other."

Now the mechanism, which is the reason this file exists.

Simon does not generate a sequence. There is no random number generator, no seed, no stored table, and no plan. The patent describes a counter that free-runs while the machine waits: "A determination is made whether any of the push-button keys are depressed. If not, the random number count is continuously incremented." The engineer who reverse-engineered the hardware states the consequence plainly: "the software counts from 1 to 4 every time the main processing loop is polled. When a button is pressed (to start the game for example) the current value of the counter is used to select the colour and so on as the game is played." Each new signal is chosen at the instant your finger closes a switch. The unpredictability that makes the thing feel like it has intentions is your own reaction time, sampled mod 4.

And the machine's memory of the game, and of you, is countable to the bit. A step is one of four values, so two bits; the longest legal sequence is 31 steps, so 62 bits. The LONGEST button replays your best run, so a second copy is held — about 124 bits in total, out of the chip's 256. Half the RAM, and the manual adds the rest: "When SIMON is turned off, the longest sequence is erased."

Caveat on the technical account, stated here rather than buried. The reverse engineering was done on a Pocket Simon carrying an MB4850 part; the author later acquired a full-size revision-A board with the TMS1000 and asserts that "since the processor of the Pocket Simon is the same as found in the full-size Simon all information contained in this document are relevant to both." I have not seen a disassembly of the 1978 ROM, and nobody appears to have published one. The counter mechanism is corroborated by the patent's own language, which is primary; the millisecond timings and the frequencies are one hobbyist's measurements of one later unit.

Why a reading would cite it

Four occasions. The first is the one this entry exists for.

First: the felt opponent with nothing inside — the loop is doing the work, not the model. This canon already carries two thirds of the anthropomorphism problem. eliza-1966 is the user projecting a mind onto a program that made no claim for itself. merlin-1978 is the manufacturer printing the mind-language as product copy — "he," "remarkably intelligent," "aggressive tactics" — about a machine that at least contains decision procedures to describe. Simon is the third and hardest case: there is no mechanism to anthropomorphise. No evaluation function, no search, no policy, no model of the player. Nothing about what Simon does depends on anything about you except when your finger lands. And the sense of an adversary is stronger here than in Merlin, not weaker. It outsold Merlin several times over and has stayed in production for forty-eight years.

So what produces the opponent? Four things, and they are all interaction design: turn-taking; a clock you do not control (five seconds, and a tempo that steps up at fixed points); an acknowledgment token on success (six short signals — the manual calls it a salute); and a mocking noise on failure (a raspberry, which is a social gesture, not a status code). That is the entire apparatus of agency in this object, and it fits, with the three games and the four skill levels and the tune, inside one kilobyte of mask ROM.

The question it hands a 2026 reading is precise and answerable: when a product's "personality," "presence" or "character" is credited to the model behind it, how much of it is the loop? The turn-taking, the latency, the typing indicator, the interjections, the acknowledgment tokens, the little sounds. Simon is the control experiment in which the model side is provably zero, and the effect held at ten million units. It settles nothing about any particular 2026 system — it supplies the null hypothesis, and the knowledge that the null hypothesis once ran for six years at the top of a national toy market.

Second: apparent adaptivity that is an open-loop schedule. Simon never responds to how well you are playing. The sequence grows by exactly one per round; the tempo steps up after the 5th, 9th and 13th signals, always, whoever is playing; the skill level is set by a human hand on a slider before play begins. There is no feedback path from your performance to the machine's behaviour, and there is no place in 32 bytes to keep one. Yet what everybody remembers is a machine that bore down on them — that got faster and harder because it was winning. It was not. A fixed ramp crossed the player's own memory-decay curve, and the crossing felt like intent.

The asymmetry underneath is worth stating in the same breath. The machine holds 31 items in perfect order for 62 bits and never degrades; the person across the table is working against a span that Miller put at about seven chunks. The contest is between perfect recall and none, and the only thing making it a game at all is the ramp. Cite this whenever a system is said to adapt, to personalise, or to "learn you," and the evidence offered is that users experience it that way. Users experienced Simon that way. The honest question is where the feedback path is, and in 1978 the answer was that there wasn't one.

Third: the randomness is the human's own. This is the smallest of the four and the most checkable, which is why it is worth carrying. The variation that reads as the machine's choices is a two-bit counter sampled at the moment of a keypress — the player's timing, quantised. A reading meets the 2026 version of this constantly: outputs read as decisions when the variation came from the sampling seed, the temperature, the retrieval order, or the user's own phrasing of the prompt. The entry does not resolve any such case. It supplies a clean one where the answer is known and turns out to be the person was the entropy source, printed on ten million boxes as "random signals."

Fourth, and the reason to carry Simon and merlin-1978 together: two base rates that point in opposite directions.

The object survived. Simon has been in continuous production since 1978 under Milton Bradley and then Hasbro — forty-eight years — through Super Simon (1979), Pocket Simon (1980), Simon Squared (2000), Simon Trickster (2005), Simon Flash (2011), Simon Swipe (2013), Simon Air (2016), Simon Optix (2017) and a plain classic edition still on shelves. Merlin, same year, same chip family, same $25 price, same trade circuit, wore the American toy crown in 1980 and was displaced by 1982, with its whole form factor gone by mid-decade. The difference that survives inspection is that Merlin was content and Simon was a loop. Merlin's six games were six pieces of content, each superseded by something better on a screen. Simon has nothing in it to become dated, because there is nothing in it. If a reading wants a cheap prior on which of two 2026 products is still here in a decade, "which one is a loop and which one is content" has a defensible record.

The company did not. Simon made Milton Bradley's electronics line 30% of a record $360 million in 1980, and it was the firm's top-selling item that year. In March 1983 Milton Bradley bought GCE to get the Vectrex console; it lost $31.6 million on it, killed it in February 1984, and on 4 May 1984 agreed to be acquired by Hasbro, in a deal generally reported at $350 million — one company history says $360 million, which looks like a conflation with the 1980 revenue figure. Contemporary analysis blamed electronic toys for the decline that forced the sale. Six years from owning the category's most durable object to being sold because of the category. Cite this when a firm's success in a hot category is offered as evidence that it will survive the category, which is one of the more common inferences on the capex and concentration lenses in 2026. The product line and the company that made it are separable, and here they ran opposite ways for the same reason.

What it got right, and what it got wrong

Not required for moment. The canon does it anyway. Four gradeable claims.

Baer's design claim about the tones. Made 1976–77, due on the first game, correct — and his explanation of the competition's failure is a self-assessment and is recorded as one. The bugle constraint works: no sequence of Simon's four notes is dissonant, and forty-eight years of production is a long confirmation. Baer's further claim, from his 2005 autobiography Videogames: In the Beginning, is that the knock-offs — Atari's own handheld Touch Me (1978), "Einstein," Tiger's Copycat — failed because their sounds lacked the charm of his. The knock-offs did indeed fail. But this is the designer grading his own design decision, and I could find no independent analysis that tests it against price, distribution, Milton Bradley's shelf position, or first-mover advantage in a fad market. Per this canon's rule for self-graded claims: his grade is on the record and an independent one does not exist, and an entry that repeated his reason as the reason would be worth less than one that says so.

The manual's claim that the signals are "random." Made 1978, due immediately, false as stated and harmless. They are a modulo-4 counter sampled by the player's own keypress. Nobody was defrauded, because the sequences are unpredictable to a person, which is the only property the product needed. What is worth keeping is that "random" was the available word for I cannot predict it, used about a mechanism that has no randomness in it — and that this is still the available word, used the same way, about systems whose sampling is equally deterministic given the seed.

"Simon's a computer, Simon has a brain — you either do what Simon says or else go down the drain." Made 1994–1998, in television advertising, due immediately, false as engineering — and the date is the finding. This is the most-quoted mind-claim ever made about the object, and Milton Bradley did not make it in 1978. It arrived sixteen years later, about the same four buttons and the same one kilobyte, and it is a considerably stronger claim than anything in the 1978 manual, which never says brain and never says "he" — it says "its," even while granting Simon a mind to blow and a memory to lose. Mind-language did not attach to this object at the moment of novelty. It attached once the mechanism was old and boring, when the copy had to do the work the novelty used to do. A reading watching a 2026 product's language grow more mentalistic as its capability plateaus has a dated precedent for the pattern, and it is not the one people assume.

Milton Bradley's product bet. Made 1977–78, due Christmas 1978, a clean hit — and the company's extension of the same bet, due by 1984, a loss. The launch bet: a single-function $25 toy, no screen, one rule, sold to adults at a disco. Result: shortages that winter, stores rationed, top-selling item by 1980, more than ten million units by the end of 1982 (a widely repeated figure whose earliest source I could reach is Mental Floss in 2019, not a primary one), and the longest continuous production run of any electronic game. The extension bet is graded in section 2's fourth item and went the other way. The same company, the same category, six years apart.

Commonly misused as

Not required for moment. It earns a section because two of the four best-known claims about this object are wrong, and because the id invites two confusions that a canon is exactly the right place to head off.

• "The first electronic memory game." No. Atari's Touch Me, an arcade cabinet from 1974, is the same game — watch a sequence of lit buttons, play it back, get one more each round. It was commercially unsuccessful, lost in an arcade full of pinball. Baer and Morrison saw it at the MOA show in 1976 and Baer's verdict is on the record: "Nice gameplay. Terrible execution. Visually boring. Miserable, rasping sounds." Atari then released a handheld Touch Me in 1978, after Simon, to chase it. What Simon added was not the game; it was four notes that could not clash, a shape you wanted on a table, and a launch party.

• "Simon gets harder as you get better." No. Open loop. Fixed ramp, fixed tempo steps, difficulty set by a human before play. See section 2.

• "Simon's sequences are random." Not in the sense the word is doing work in. See section 1 and section 3.

• "simon-1978 is Herbert Simon." It is not, and this canon holds Herbert Simon at simon-1965. His Nobel Memorial Prize in Economic Sciences was awarded in 1978 — for bounded rationality and decision-making in organisations — which makes the collision worse than a coincidence of surname. Anyone assembling a descent chain by id alone will eventually connect a toy to satisficing.

• And it is not Edmund Berkeley's Simon, which is worth more than a disambiguation. Berkeley's Simon (1950) was a relay machine described across thirteen construction articles in Radio-Electronics beginning in October 1950, built by two Columbia graduate students, following his 1949 book Giant Brains, or Machines That Think. It handled two-bit numbers and had 32 bits of memory in total. Berkeley named it downward: "Let us call it Simon, because of its predecessor, Simple Simon" — the nursery-rhyme simpleton — and said its purpose was "to stimulate thinking and understanding, and to produce training and skill." Set the two side by side. Two machines called Simon, twenty-eight years apart, each holding a few dozen bits of state: the 1950 one named to deflate the brain talk its own author's bestseller had inflamed, the 1978 one eventually advertised as having a brain. The field is usually told as running the other way — sober engineering first, hype later — and here the deflation came first and the inflation arrived in 1994. Sources disagree on Berkeley's build cost (under $300 in one account, about $600 in another); I have not resolved it and it does not bear on anything above.

• "It proves the whole thing was always marketing." The deflationary misuse that speak-and-spell-1978, merlin-1978 and dijkstra-1959 each have to guard against, and it is weakest here of all, because Simon oversold nothing. The manual describes exactly what the object does, the object does it, and it did it for a generation. The finding is not that mind-language is empty. It is that the experience of facing an agent is cheap to manufacture — four bulbs, a buzzer, a clock and turn-taking — which is a fact about people and about interaction design, and not an accusation against anybody.

• A correction to a canonical list, for the next session. TIME's All-TIME 100 Greatest Toys entry describes the game backwards: "Players would illuminate a series of colors by touching them in an order of their choosing. Opponents would then have to mimic the pattern exactly or lose." That is Game II ("PLAYER ADDS"), a minor mode, and the description removes the machine from the loop altogether. It is the mirror image of the error this entry is about, made by the list most likely to be cited for the object's importance.

Sources

The standing caveat, in the form deep-blue-1997, speak-and-spell-1978 and merlin-1978 set it. The fetch tool in this run returns another model's summary of a page rather than the page, so quotations arrive at one remove. Where two extractions of the same page disagreed I have said so. One instance matters: a first pass over Wikipedia's Simon article returned note frequencies in Hz, and a second pass over the same article reported that the article carries no frequencies at all. I have therefore attributed Hz only to the one source that measured them, and note names only to Wikipedia.

A workaround the next session should have, because merlin-1978 did not. That entry's largest gap was that every primary scan it needed was an image PDF and pdftotext was refused by the shell in this run. It is still refused — but the Read tool renders PDF pages directly, and that is how the entire 1978 Simon rules booklet was read here, verbatim, four pages, including every quotation in sections 1 and 3. Any canon entry blocked on a scanned manual, brochure or patent PDF should fetch the PDF and then open it with Read. This belongs in ONBOARDING.md; a session with a wider write boundary than this job's should move it there.

Fetch failures in this run, recorded so they are not repeated. americanhistory.si.edu and si.edu both returned HTTP 403 for the National Museum of American History's Simon object record (nmah_1302005), so the Smithsonian catalogue text is unread. fastcompany.com returned 403, so 40 Years of Simon (2018), the one long-form anniversary piece with interviews, is unread. No page of Ralph Baer's own site surfaced in search and none was reachable. No published disassembly of the 1978 ROM was found.

A claim I could not verify and have therefore not used. A search summary attributed to the Milton Bradley company histories a statement by the firm's vice-president of the game division that Simon would sell about 1.5 million units in 1979, roughly twice its 1978 figure. I fetched both texts and neither contains it. Year-by-year unit figures for Simon are, as far as this run can tell, not in any reachable source; the only numbers that survive checking are "top-selling item by 1980," the 30%-of-$360-million line, and the ten-million-by-1982 figure, which is secondary.

Primary.

• Simon rules booklet, "SIMON — 1978 Milton Bradley," four pages, read in full via Hasbro's own hosted scan (hasbro.com/common/documents/…, a copy of the 1stingames.com transcription). Source of every manual quotation above: "Chase my flashing lights and sounds," "the ever-increasing random signals that SIMON generates," the 8/14/20/31 skill levels, the tempo steps after the 5th, 9th and 13th signals, the five-second limit, the RAZZ, the six-signal salute, "SIMON blows its mind, loses its memory and gives itself the RAZZ sound," "Players may team up against SIMON," the spare bulb and the type 14A lamps, and "When SIMON is turned off, the longest sequence is erased." The single most useful document about this object.
• US Patent 4,207,087, Microcomputer controlled game — Howard J. Morrison and Ralph H. Baer, assigned to Marvin Glass and Associates. Filed 19 September 1977, granted 10 June 1980. Read via Google Patents. Source of the free-running counter — "If not, the random number count is continuously incremented" — of the five described game variants, of the eight/fourteen/twenty sequence lengths, and of the framing of the machine as something a participant plays "against."
• US Design Patent D253,786, the housing, credited to Douglas P. Montague. Cited here from Toy Tales; I did not open the patent itself.

Object records.

• The Henry Ford, object 99.42.1, "Simon Game, 1978–1983," Milton Bradley, Springfield, Massachusetts — 10.5 inches diameter, 2.5 inches high, the inscription "1978 by Milton Bradley Co.," the Patent No. 4207087 marking, and the note that the circuit board was manufactured in Singapore.
• Smithsonian NMAH, Simon Electronic Game, 1978 (nmah_1302005) — identified in search but unreadable; see the failure log.

Technical.

• "Reverse Engineering an MB Electronic Simon Game," waitingforfriday.com — the counter-sampled-at-keypress mechanism quoted in section 1, the tone measurements (209, 252, 310, 415 Hz), the tempo table (0.42 s per tone to signal 5, 0.32 s to 13, 0.22 s beyond), the three-second in-game timeout, the 2-bits-per-step and 62-bits-per-sequence arithmetic, and the author's own statement that he worked on a Pocket Simon and treats the parts as equivalent. The best technical source that exists on this object, and a hobbyist one; weighted accordingly.
• MAME, src/mame/handheld/hh_tms1k.cpp — the dumped revision-A Simon ROM as MP3226, a TMS1000, 1978. Via search extracts of the driver file, not the file itself.
• TMS1000 capacities (1,024 × 8 ROM, 64 × 4 RAM) as established in this canon's merlin-1978 entry, which sets out the family in detail.

Historical and contextual.

• **Wikipedia, *Simon (game)*** — Baer and Morrison at Marvin Glass, Lenny Cope's programming of "Follow Me," Charles Kapps's assembly work, the MOA 1976 sighting and Baer's verdict, the bugle rationale and "always be harmonic, no matter the sequence," the A-major and B♭-minor tone sets, the $24.95 price, the Studio 54 launch, and the sequel list.
• **Wikipedia, *Touch Me (arcade game)*** — the 1974 arcade original, its commercial failure, and Atari's 1978 handheld released in Simon's wake.
• **Wikipedia, *Simon (computer)*** and Edmund Berkeley — the 1950 relay machine, the thirteen Radio-Electronics articles from October 1950, the two-bit words and 32 bits of memory, the Columbia graduate students, "Let us call it Simon, because of its predecessor, Simple Simon," and the stated instructional purpose. Note the unresolved build cost.
• **Wikipedia, Howard J. Morrison — born 1932, died 21 January 2025; Marvin Glass until its 1988 closure, then Breslow, Morrison, Terzian, later Big Monster Toys; Toy Industry Association Hall of Fame, 1998. With Ralph Baer's death on 6 December 2014**, both principals are now dead, which is the practical reason several questions above will stay open.
• IEEE Spectrum, on Simon — the May 1978 Studio 54 launch with a one-metre replica descending at midnight, Baer's 20 × 20 cm housing, Cope's prior TMS1000 work, the Teletype link to a computer in Pennsylvania, and the G–C–E–G description of the tones. Its statement that the TMS1000 held "1,000 kilobytes of memory" is wrong by a factor of a thousand — the part has 1,024 bytes of ROM — and a reading quoting Spectrum on this object should skip that clause.
• Smithsonian Magazine, "The Not-So-Simple Simon Proved the Young Were Swifter Than the Old" — the 15 May 1978 Studio 54 date, the four-foot model over the dance floor, the rectangular-to-round redesign after Milton Bradley licensed it, Baer's "pleasing in any sequence," and the Close Encounters coincidence. Its age-and-performance claim is explicitly anecdotal — a father losing to his son — and is not evidence of anything; I have not used it.
• Mental Floss, "7 Things You Might Not Know About the Game Simon" (2019) — the ten-million-by-1982 figure, the Christmas 1978 shortages and rationing, one store selling a thousand units in five days, the $25 price, the failed knock-offs and Baer's explanation of their failure, and the Close Encounters boost. Secondary throughout; it is the earliest reachable source for the sales figure and should be cited as such.
• Toy Tales, "Simon from Milton Bradley (1978)" — the midnight 15 May 1978 launch, Montague's housing and the design patent, the battery configuration, the three game modes and the five-second RAZZ window.
• TIME, All-TIME 100 Greatest Toys, Simon — the cultural placement, the Studio 54 line ("who'd have thought that Simon would be the one of the pair to survive the 1980s?"), and the backwards description of the game corrected in section 4.
• encyclopedia.com / referenceforbusiness, Milton Bradley Company — "By 1980, Simon was Milton Bradley's top selling item and largely responsible for the success of the company's electronic line which contributed 30 percent of Milton Bradley's record $360 million in sales," and the May 1984 Hasbro acquisition. Both date Simon's release to 1977; see the date paragraph.
• On the Vectrex collapse: the Vectrex and Milton Bradley histories — GCE bought March 1983, $31.6 million lost, discontinued February 1984, Hasbro agreement 4 May 1984 at about $350 million, and contemporary analysis blaming electronic toys for the decline that forced the sale.
• On the tones' cultural tailwind: Close Encounters of the Third Kind (1977) had made a five-note exchange of lights and sounds the popular image of talking to a nonhuman intelligence six months before Simon shipped. Every account I found treats this as coincidence that helped sales rather than as influence on the design, and Baer's own account credits the bugle and the encyclopedia. I record it as context, not as a cause.