The roulette computer, and the distance between a proven edge and a working one

Filed under idea in canon/proposals.md — the entry sits inside that file's ## idea block. That filing is defensible and I am not using it. It is defensible because, unusually, a documented descent really does exist: the field of wearable computing has formally adopted this device as its origin. IEEE's Second International Symposium on Wearable Computers invited Thorp to write the story up in 1998; Guinness World Records certifies the 1961 device as the first wearable computer; the MIT Museum holds the object. That is a lineage claim made by institutions, not by me.

It is the wrong filing anyway, and for the reason card-counting-1962 — the entry written immediately before this one, on the same day — set out about its own case. This canon's idea is "a framing later work is built out of rather than argued about," cited "when the thing being watched descends from it and the descent explains its shape." The descent here is real but it points at wearable computing, and wearable computing is not what a reading is watching. No agent framework shipping this month traces its ancestry to a switch under a big toe. What this entry supplies to a reading is precedent and analogy, and filing it as idea would have the canon assert a lineage that does not reach the subject — a reader following the claim back would arrive at a shoe.

moment — "something that happened and changed what people expected next" — is right, and the event is not the invention. It is the wearing. In the summer of 1961 a computer left a building and was operated in the field, in real time, against an adversary, by people who needed it to be invisible. The expectation that eventually changed, and which is now written into Nevada criminal law, is that the person across the table may be augmented and you cannot tell by looking.

The weak point in that argument, stated plainly. The thing was secret for five years. Nothing anyone could observe changed in 1961; the disclosure came in 1966, and the institution did not write a rule until 1985. A strict reading of moment wants the expectation to move when the event happens. I accepted the filing anyway because the interval between capability and public knowledge is one of the things this entry is for — an entry that could not be filed as a moment on the grounds that nobody knew yet would be throwing away its own best material. A reading using this entry should carry the caveat with it.

Why not the others. limit is backwards: this is a positive result about what can be done to a system that was the culture's standing symbol for what cannot be predicted. interpretation fails the same test it failed for card-counting-1962 — every interpretation in canon/ is a named, published reading with an author, and there is no named reading here. prediction fails because +44% is an expectation in the probabilistic sense, not a dated forecast about the world; it is nonetheless graded below, because it turned out to be checkable.

A note on ancestry, including the inverted year. descends_from holds card-counting-1962, and the year inversion — a 1961 entry descending from a 1962 one — is deliberate rather than a slip. The descent is not from the book. It is from the paper. A submission to the Proceedings of the National Academy of Sciences required a member of the Academy to communicate it; Thorp, then a mathematics instructor at MIT, took his blackjack work to Shannon in the autumn of 1960 for exactly that reason; Shannon communicated "A Favorable Strategy for Twenty-One" on 21 November 1960, and the roulette project began out of that meeting. The canon entry filed as card-counting-1962 covers that whole body of work, including the 1959 IBM 704 run and the January 1961 paper, so the link is to a real ancestor under a later headline year.

Two ancestors a reader will want are not in canon/ and should not be invented. shannon-limit-1948 is proposed in canon/proposals.md and unwritten. J. L. Kelly's 1956 growth-optimal bet size — the thing that converts an edge into a bankroll that survives its own variance — is proposed nowhere and is load-bearing for both Thorp entries. shannon-chess-1950 is written, and is the same Claude Shannon, and is not an ancestor of this; a reading should link the two as the same man, not as a lineage.

What it is

Roulette is the game the culture uses when it wants a picture of pure chance. A croupier spins a wheel one way and releases a ball around a track the other way; the ball loses energy, drops, rattles and settles into one of the numbered pockets — 38 of them on the American wheel, which includes both a zero and a double zero and gives the house 2 in 38, or about 5.26%, of every bet made on it forever. There is no memory in the wheel, no drift to exploit, nothing to count. It is the flat, honest, unbeatable case.

It is also a mechanical system with a position, a velocity and a deceleration, and Thorp had noticed this as a schoolboy. The UCI Libraries exhibit of his papers quotes him from his 2017 memoir: "Since planetary positions were accurately predictable, I thought I might be able to forecast the outcome of a roulette pin." (That transcription almost certainly renders "spin" as "pin"; I am quoting the exhibit's text, not the book, and flag the slip rather than silently repairing a quotation.) He dates the conception of the machine to 1955.

What made the idea operational was a procedural gap, not a physical one. Casinos let players keep betting for some seconds after the ball is released. That window is the whole opportunity: it is the only interval in which a measurement of the actual spin can still be converted into a wager. Everything else in the project is engineering in service of using it.

In the autumn of 1960 Thorp met Claude Shannon at MIT. The UCI exhibit puts the meeting in September 1960; the timeline in Engadget's 2013 interview piece puts it in November; the hard date in the record is the "Communicated by Claude E. Shannon, November 21, 1960" line on Thorp's blackjack paper. The discrepancy does not matter to anything downstream and is recorded so that a later reading does not treat any one of the three as settled.

For roughly the next eight or nine months they worked in the basement lab of Shannon's house — a large three-storey wooden house on one of the Mystic Lakes, several miles from Cambridge, in Winchester, Massachusetts. Thorp described it as a gadgeteer's paradise holding something like a hundred thousand dollars' worth of electrical, electronic and mechanical parts. The accounts differ on the effort: one says Thorp gave about half his time for eight months, another that the two averaged twenty hours a week for nine. They bought a regulation roulette wheel from Reno for $1,500, a strobe light, and a clock whose hand went round once per second, and they measured the thing.

The machine that came out of it was about the size of a cigarette pack, was analog, and had twelve transistors. Wires ran down to switches in the toes of the operator's shoes: one toe-click as a reference point on the wheel passed, another as the ball passed, which is enough to give the machine both velocities. Wires ran up to a radio link built partly from model-aircraft transmitter and receiver parts, and the receiver drove a tiny loudspeaker pushed into the ear canal of a second person — the bettor — sitting at the table. The output was eight musical tones, one per octant of the wheel. The last tone you heard was where to put the chip.

It took two people, and it took a rehearsed body: Thorp and Shannon trained their big toes to work the switches. The final operating version was tested in Shannon's basement in June 1961. The laboratory result, in Thorp's own framing, was an expected gain of +44% when betting the most favoured octant — against a house edge of −5.26%. (The UCI exhibit reports the numbers slightly differently: a standard deviation of about ten pockets on a single-number prediction, converting to a 44% edge on a single number and 43% on an octant. Thorp's own abstract attaches the 44% to the octant. The retellings have swapped which number belongs to which bet; both agree the edge was somewhere around 44%.)

In the summer of 1961 — August, by the exhibit's account — the Thorps and the Shannons went to Las Vegas: four people, two of whom had to stand near a wheel with switches in their shoes while the other two sat and bet. It worked. Thorp's memoir, as quoted by the exhibit, describes "turning small piles of dime chips into large ones," and his 1998 write-up says the predictions there were consistent with the 44% figure from the lab.

And that is the end of it, for a reason worth quoting at length, because it is the reason and not a metaphor for the reason. The failure was the wire. From the memoir, via the exhibit: "Even though they were steel, they were so fine that they broke frequently, leading to long interruptions while we returned to our rooms and went through the tedious process of doing the repairs and then rewiring [the bettor]." And the other failure was the body the machine had to hide inside: "Well into one winning session, a lady next to me looked over in horror. Knowing I should leave, but not why, I raced to the restroom and there in the mirror saw the speaker peeking out from my ear canal like an alien insect." (Thorp told Engadget a shorter version of the same story in 2013 — "Once a lady next to me looked over in horror. I left the table quickly and discovered the speaker peering from my ear canal like an alien insect." The two differ in wording; neither is a transcript of the other.)

They stopped. They also kept quiet: the existence of the computer was secret until Thorp described it in the 1966 revised edition of Beat the Dealer, and the technical account followed in the Review of the International Statistical Institute in 1969. A related prediction system for the Wheel of Fortune had appeared earlier, in Life, 27 March 1964. Thorp finally wrote the invention up as history in 1998, at the wearable-computing field's own conference, under the title "The Invention of the First Wearable Computer." The device is in the MIT Museum.

Why a reading would cite it

The primary case: the gap between a validated capability and a working one, measured in a case where both ends are documented. The laboratory number was +44%. The field result was dime chips. Nothing about the model was wrong — it has since been independently reproduced, twice, by people who had no stake in defending it — and the project still produced no money, because fine steel wire breaks, because a repair means leaving the floor, because a loudspeaker works its way out of an ear, and because two frightened mathematicians and their wives are not an operation. Cite this when a reading is grading a capability claim whose evidence is a benchmark, a demo or an internal eval. The question that separates 44% from dime chips is not "does the method work." It is whether the apparatus survives eight hours of contact with an unreliable, adversarial, physical world, run by people who can be seen. That is the specific gap that most 2026 claims about autonomous agents are sitting on, and this is the case where the two numbers are both on the record and the distance between them has a cause you can name.

Second: where a new computational capability goes first, and where it goes next. Wearable computing was invented at least three separate times before it had a name, and all three times it was invented to take money out of a casino. Thorp and Shannon, 1960–61, analog, roulette. Keith Taft's blackjack computers through the 1970s and 1980s. And the Eudaemons — Doyne Farmer, Norman Packard and others at UC Santa Cruz in the late 1970s — who built what Farmer calls the first wearable digital computer, a three-kilobyte hand-coded machine-language program with toe inputs and vibrating outputs, made over eleven trips to Las Vegas, Reno and Tahoe, and by their own account achieved a 20% advantage. Then watch the second move: Thorp went to the markets, founding Convertible Hedge Associates with Jay Regan in November 1969, renamed Princeton/Newport Partners in 1974. Farmer, Packard and James McGill founded Prediction Company in Santa Fe in March 1991; it contracted exclusively to O'Connor and Associates in September 1992, went with O'Connor into Swiss Bank Corporation and thence into UBS, and UBS bought it outright in 2005. Cite this when characterising where a capability lands first: not because casinos are special, but because the earliest sustained deployment tends to be wherever the payoff is immediate, measurable and denominated in cash — and because the same people then carry the method to where the money is larger and the rules are thinner. Three cases is a pattern and not a base rate, and a reading should say so rather than dress it as one.

Third: how long the rule takes, what triggers it, and how wide it ends up. In 1961 Nevada had no law about computers in casinos. The first one is NRS 465.075, added by the Nevada legislature in 1985 — twenty-four years later. Two things about it are the citable part. It was not triggered by this device or by anything like it: the reported proximate cause was one of Keith Taft's machines, a miniature camera in a belt buckle relaying the dealer's hole card out to a receiver in a truck. And once written, it kept being widened. Its section heading in the year-stamped code reads "Use of device for calculating probabilities" in 2010; "Use or possession of device to obtain advantage at playing game in licensed gaming establishment" in 2011; and today "Use or possession of device, software or hardware to obtain advantage at playing game prohibited." The current text reaches "any computerized, electronic, electrical or mechanical device, or any software or hardware, or any combination thereof, which is designed, constructed, altered or programmed to obtain an advantage at playing any game in a licensed gaming establishment or any game that is offered by a licensee or affiliate," and enumerates, among others, a device that "Projects the outcome of the game" and one that "Analyzes the strategy for playing or betting to be used in the game." A first violation is a category C felony. Cite this when a reading is describing the lag between a capability and a rule, or the shape a late rule takes: written decades after the capability, in response to a different instance of it, then amended (2011, 2013) until it covers an object every customer is already carrying. Fifty-two years separate a wheel in a Winchester basement from a statute that reaches an app on a phone.

Fourth: augmentation, and where the answerability lands. The machine never played. It computed an octant and whispered it to a person, who decided whether to place a chip. The law that eventually arrived does not regulate the device as a device; it makes a person a felon for using one, or for possessing one with intent to use it. Cite this when a reading needs to be precise about who is answerable for a machine-assisted act — the shape here is not "the machine did it" and not "the person did it unaided," and both the 1961 operation and the 1985 statute agree on which of the two the world holds responsible.

What it got right, and what it got wrong

Not required for a moment, but there are dated, checkable claims here and leaving them ungraded would waste the entry.

The physics: right, and independently confirmed twice by people with no stake in it. The claim, made publicly in 1966 and in technical form in 1969, is that a physical roulette wheel is deterministic enough that a few timing measurements taken during the betting window convert a negative expectation into a large positive one. The Eudaemons reproduced it in the late 1970s at a self-reported 20% — same team's own account, so treat it as interested. The independent check is Michael Small and Chi Kong Tse, "Predicting the outcome of roulette," Chaos 22, 033150 (2012), submitted 28 April 2012. Their abstract, verbatim: they "provide a very simple model for the motion of a roulette wheel and ball and demonstrate that knowledge of initial position, velocity and acceleration is sufficient to predict the outcome with adequate certainty to achieve a positive expected return," and with a rig relying "only on a mechanical count of rotation of the ball and the wheel" they "demonstrate an expected return of at least 18%, well above the -2.7% expected of a random bet." That is a peer-reviewed physics journal, fifty-one years later, with a cruder instrument than Shannon built, on the more favourable European wheel, getting a number of the same kind. The 1961 claim is as well supported as an old private claim gets.

The business: wrong, and Thorp says so himself. The project made no meaningful money. The entire recorded yield is dime chips, and it was abandoned in its first week of field use. Everything Thorp actually earned came from the method that needed no device — blackjack — and then from the markets. A reading that cites this entry as a story about a machine that beat a casino has taken the wrong half. It is a story about a machine that could have beaten a casino and did not, and the causes were mechanical, physiological and organisational rather than mathematical.

The stated reason for quitting: partly right, by the wrong mechanism. Thorp gives, alongside the hardware, the knowledge that casinos could stop allowing bets after the ball was released at any time. The countermeasure that actually arrived, twenty-four years on, was not procedural but criminal: a felony statute. I have no source establishing what betting windows look like on a 2026 floor and will not assert one. What can be graded is the form of the response, and the form was law, not house rules — the opposite way round from card-counting-1962, where the house changed its own game because no law could be written against a number held in someone's head. A device can be legislated against. A thought cannot. That contrast is the reason these two entries are worth holding together.

The "first wearable computer" claim: upheld, with a caveat the field itself supplies. Made by Thorp in 1998, adjudicated since by Guinness World Records (holders "Edward O Thorp and Claude Shannon," 1961) and by the wearable computing community that invited him to make it. The caveat is that the 1961 device was analog, single-purpose, and needed two operators and a radio link; Farmer describes the Eudaemons' shoe machine of the late 1970s as the first wearable digital computer. Both claims can be true and a reading should carry the adjective.

The migration, graded on its own terms. Thorp's fund is the most-cited proof that the casino methods travelled, and the numbers in circulation for it do not agree: one account gives roughly 20% annualised after fees for over two decades with no down quarter, another 19 years of gains with only three losing months. Both are secondary and I could not reach an audited figure; a reading quoting either should quote it as a claim. What is documented is the ending, and the ending complicates the "capability moves to where the rules are thinner" story rather than confirming it. About fifty armed marshals raided Princeton/Newport in December 1987; six people were convicted on 1 August 1989; the Second Circuit reversed on 28 June 1991 in United States v. Regan, 937 F.2d 823. Edward Thorp is not among the six named defendants in that caption. The firm was already gone — wound down from late 1988, closed in 1989 — killed by the cost of the prosecution rather than by its outcome. The rules were not thinner. They were differently shaped, and enforced by an institution that did not need a conviction to close a business.

Shannon's own record, flagged rather than repeated. The widely circulated figure is that Shannon's stock portfolio returned about 28% a year from the late 1950s to 1986. It traces to Poundstone's Fortune's Formula and ultimately to Thorp's recollection; it is not audited and this entry does not endorse it. It is recorded here only because a reading tempted to use it should know it is a friend's number about a dead man's private accounts.

Commonly misused as

Not required for a moment, and included because every one of these has been seen in circulation.

"Thorp and Shannon beat roulette." They demonstrated an edge and never converted it. The gap between demonstrating and converting is the entire value of the entry, and this phrasing deletes it.

"The first wearable computer" — meaning the first computer worn. It was an analog, single-purpose timing and prediction circuit of twelve transistors, requiring a second operator, a radio link and a rehearsed toe. It is the first wearable computer in the sense the field has agreed to; it is not a small general-purpose machine, and a reading that uses it to date the ancestry of anything programmable is overreaching. The first wearable digital computer came about sixteen years later and was also built to beat roulette.

"Roulette is beatable, so any system can be beaten." What is established is narrow and physical: a mechanical wheel is a deterministic system, and a sufficient betting window plus adequate instruments makes its outcome predictable well enough to profit. It establishes nothing about an RNG-driven electronic wheel, nothing about a game with no physical state to measure, and nothing whatever about markets — Thorp's later work is a separate argument that stands or falls on its own evidence. Small and Tse also found that a slight slant in the table produces a pronounced bias, which is a fact about a particular table, not about chance.

"Nevada banned these devices because of Thorp." No. Twenty-four years separate the device from the statute, and the reported trigger was a different machine entirely. A reading wanting an example of regulators reacting to a capability should use this as the slow case, which is what makes it worth having.

"It shows regulation catches up fast." It shows the reverse, twice over: twenty-four years to the first rule, and then two further amendments across another twenty-eight years to keep the rule pointed at the capability as the capability changed substrate.

"They were caught." They were not. A woman looked at Thorp oddly, an earpiece worked loose, and they went home. Nobody was detected, accused or excluded. The project ended because the hardware was not good enough and the operators were frightened, which is a much more ordinary and much more transferable ending.

Sources

Primary, bibliographic record only — the paper could not be read from here. Edward O. Thorp, "The Invention of the First Wearable Computer," Digest of Papers. Second International Symposium on Wearable Computers (Cat. No. 98EX215), pp. 4–8, DOI 10.1109/ISWC.1998.729523. The venue string, the single-author record and the publication date of 19 October 1998 were read from Semantic Scholar's API record for that DOI (DBLP key conf/iswc/Thorp98), which explicitly notes that the abstract field has been elided by the publisher. Copies of the PDF at cs.virginia.edu and cs.columbia.edu were reached but are 1998 Distiller output with no extractable text layer, and no text-extraction tool was available to this job; IEEE Xplore, the ACM Digital Library and ResearchGate returned nothing usable or 403. Nothing in this entry is quoted from the body of that paper. The sentences attributed to its abstract in the prose above — the 1955 conception, the June 1961 basement test, the +44% on the favoured octant, the summer 1961 Las Vegas trip being consistent with the laboratory figure, and the secrecy until 1966 — were read from search-engine reproductions of the abstract, and should be treated as a secondary rendering of a primary text rather than as the text.

Primary, read at source. Nevada Legislature, NRS Chapter 465, at leg.state.nv.us. Source of the verbatim current text of NRS 465.075 quoted above, of its history note "(Added to NRS by 1985, 970; A 2011, 216; 2013, 1317)", and of NRS 465.088's category C felony for a first offence, itself noted as "(Added to NRS by 1981, 1292; A 2025, 2282)".

Primary, bibliographic and via search-result page titles only. Justia's year-stamped Nevada codes for NRS 465.075. Justia returned 403 to direct fetches from here, so the three section headings quoted above — the 2010 "Use of device for calculating probabilities," the 2011 "Use or possession of device to obtain advantage at playing game in licensed gaming establishment," and the current "Use or possession of device, software or hardware to obtain advantage at playing game prohibited" — come from the titles Justia serves for those year-stamped pages, not from the page bodies. The current wording is independently confirmed by the legislature's own site above; the two earlier headings are not.

Primary, read at source. arXiv:1204.6412, Michael Small and Chi Kong Tse, "Predicting the outcome of roulette," submitted 28 April 2012, revised 13 July 2012, published as Chaos 22, 033150 (2012), DOI 10.1063/1.4753920. The abstract is quoted verbatim above and is the source of the 18% expected return, the −2.7% baseline for a random bet on a European wheel, the mechanical-count rig, and the table-slant finding.

Primary, read at source. J. Doyne Farmer's own roulette page at doynefarmer.com. Source of the Eudaemons' account in Farmer's words: the first wearable digital computer, the three-kilobyte hand-coded machine-language program, the floating-point package and sequencer, toe inputs and vibrating outputs, the toe switch used to measure ball and rotor, the ten-second window between release and close of betting, the eleven-plus trips to Las Vegas, Reno and Tahoe, the 20% advantage, the technical failures, and the shoe computer's current loan to the Heinz Nixdorf Museum.

Primary-archival, read at source. UCI Libraries, "Finding the Edge: The Work and Insights of Edward O. Thorp," exhibits.lib.uci.edu/thorp, the "A Spin of the Wheel" page. Source of the roulette explanation and the 38 pockets, the September 1960 meeting date, the schoolboy planetary analogy quoted from A Man for All Markets p. 44, the two-person operation and the trained big toes, the standard-deviation-of-ten-pockets and 44%/43% figures, the August 1961 Las Vegas date, and the three passages quoted above from A Man for All Markets p. 133 (the dime chips, the broken steel wires, the speaker in the ear canal). It is also the source of the 1985 legislative citation — Senate Bill 467, 1985 Nevada State Senate, and "Blackjack Devices to Count Cards Barred by Nevada Assembly," Los Angeles Times, 24 May 1985 — and of the MIT wearable-computing timeline entry it reproduces (Bradley Rhodes, "A Brief History of Wearable Computing"), which is where the Life, 27 March 1964, pp. 80–91 Wheel of Fortune disclosure and the Review of the International Statistical Institute 37(3), 1969 technical publication come from. **I did not read A Man for All Markets.** Every quotation from it in this entry is quoted from this exhibit's transcription, page numbers included, and one of them contains an apparent transcription error ("roulette pin"), which is why the distinction matters. Note also that the exhibit's live path is exhibits.lib.uci.edu, not the lib.uci.edu/online-exhibits path search engines still return — the same trap card-counting-1962 recorded.

Secondary, read at source. Donald Melanson, "Gaming the system: Edward Thorp and the wearable computer that beat Vegas," Engadget, 18 September 2013, built on an interview with Thorp. Source of the twelve transistors, the model-aircraft transmitter and receiver parts, the eight tones for eight octants, the November 1960 meeting date that disagrees with the exhibit's September, and the shorter 2013 telling of the ear-canal story.

Secondary, read at source. "Ed Thorp, Claude Shannon and the world's first wearable computer," Winton, 12 October 2017. Source of the twenty-hours-a-week -for-nine-months figure, the four-person Las Vegas party (Betty Shannon and Vivian Thorp), the "19 years of gains with only three months of losses" figure for Princeton/Newport, and the 2¹¹ = 2048 blackboard anecdote. A fund manager's house publication writing about a famous fund manager; treat the performance number accordingly.

Secondary, read at source. Guinness World Records, "First wearable computer" — holders Edward O Thorp and Claude Shannon, 1961, MIT, device now at MIT. Used only as evidence that an outside body adjudicates the priority claim, not as evidence about the device.

Secondary, read at source. Wikipedia's Princeton Newport Partners article, for the November 1969 founding as Convertible Hedge Associates with Jay Regan, the 1974 renaming, the "annualized rate of return of 20 percent after fees for over two decades, without a single down quarter" claim, the liquidation under RICO costs, and the TGS Management successor. Its cited but unread references are the source of the 1 August 1989 New York Times "Six Guilty Of Stock Conspiracy," the 7 November 1989 Los Angeles Times sentencing report, and the 29 June 1991 Los Angeles Times report of the reversal. None of those three newspaper pieces was read at source.

Primary, read at source via search API. CourtListener's opinion search for United States v. Regan — source of the citation 937 F.2d 823, the Second Circuit, the filing date of 28 June 1991, the docket numbers 89-1591, 89-1592, 89-1600 to 89-1602 and 89-1614, and the full caption naming James Sutton Regan, Jack Z. Rabinowitz, Steven Barry Smotrich, Charles M. Zarzecki, Paul A. Berkman and Bruce Lee Newberg. That caption is the entire basis for this entry's statement that Thorp is not among the defendants; the opinion text itself was not read.

Secondary, weakly sourced, flagged as such. The claim that the 1985 Nevada statute was triggered by Keith Taft's belt-buckle hole-card camera, found on his brother Ted, comes from gambling-industry web sources (a Blackjack Hall of Fame profile and similar) and was not verified against the 1985 legislative record or the Los Angeles Times report. It is reported here as a reported cause. The 1985 date of the statute itself does not depend on it and is primary.

Secondary, unverified, recorded only to be discounted. The ~28% annual return attributed to Claude Shannon's personal portfolio from the late 1950s to 1986, and the claim that this beat 1,025 of the 1,026 mutual funds Barron's reported in 1986, trace to William Poundstone's Fortune's Formula (2005) and were read only in secondary summaries of it. Not audited, not endorsed.

Reached and unusable. PubMed Central's copy of Thorp's 1961 PNAS paper (PMC285252) returned a reCAPTCHA gate on this run, so the "Communicated by Claude E. Shannon, November 21, 1960" line — the hinge of the ancestry argument above — is taken from a search-engine reproduction of that front matter rather than read at source. Justia, the ACM Digital Library, ResearchGate and IEEE Xplore returned 403; the MIT Museum's record for the device (accession 2007.030.014) returned HTTP 429 on three attempts across this run, so nothing about the object's catalogue description, materials or credit line is claimed here beyond the fact, from UCI and Guinness, that the museum holds it.