Have you ever tried programming with a language that uses musical notation? What about a language that never runs programs the same way? What about a language where you write code with photographs? All exist, among many others, in the world of esoteric programming languages, and Daniel Temkin has written a forthcoming book covering 44 of them, some of which exist and are usable to some interpretation of the word “usable.” The book, Forty-Four Esolangs: The Art of Esoteric Code, is out on 23 September, published by MIT Press. I was introduced to Temkin’s work at the yearly Free and Open source Software Developer’s European Meeting (FOSDEM) event in Brussels in February. FOSDEM is typically full of strange and wonderful talks, where the open-source world gets to show its more unusual side. In Temkin’s talk, which I later described to a friend as “the most FOSDEM talk of 2025,” he demonstrated Valence, a programming language that uses eight ancient Greek measuring and numeric symbols. Temkin’s intention with Valence was to emulate the same ambiguity that human language has. This is the complete opposite of most programming languages, where syntax typically tries to be explicit and unambiguous. “Just as you could create an English sentence like, ‘Bob saw the group with the telescope,’ and you can’t quite be sure of whether it’s Bob who has the telescope and he’s seeing the group through it, or if it’s the group that has the telescope,” he says. “What if we wrote code that way so you could write something, and now you have two potential programs? One where Bob has a telescope and one where the group has a telescope.” How Esoteric Languages Spark Creativity Creating a language or an interpreter has often been the proving ground of many engineers and programmers, and esoteric languages are almost as old as non-esoteric ones. Temkin says his current effort has a lot to do with AI-generated code that seeks to do nothing but provide seemingly straight solutions to problems, removing any sense of creativity. Esoteric languages inherently make little sense and frequently serve little purpose, making them conceptually completely counter to AI-generated code and thus often not even understood by them—almost the code equivalent of wearing clothing to confuse facial recognition software. While the syntax of esoteric languages may be hard to understand, the actual programming stack is often wonderfully simple. Temkin believes that part of the appeal is also to explore the complexity of modern programming. “I come back a lot to an essay by Joseph Weizenbaum, the creator of the Eliza Chatbot, about compulsiveness and code,” he says. “He described ‘the computer bomb,’ the person who writes code and becomes obsessed with getting everything perfect, but it doesn’t work the way they want. The computer is under their control. It’s doing what they’re telling it to do, but it’s not doing what they actually want it to do.” “So they make it more complicated, and then it works the way they want,” Temkin adds. “This is the classic bind in programming. We command the machine when we’re writing code, but how much control do we really have over what happens? I think that we’re now all used to the idea that much of what’s out there in terms of code is broken in some way.” Temkin explored the idea of control in his language Olympus, where the interpreter consists of a series of Greek gods, each of which will do specific things, but only if asked the right way. Temkin’s Olympus language includes an interpreter consisting of Greek gods, which must be asked to do things in the proper way.Daniel Temkin “One example regarding complicating our relationship with the machine and how much we’re in control is my language, Olympus, where code is written to please different Greek gods,” says Temkin. “The basic idea of the language is that you write in pseudo-natural language style, asking various Greek gods to construct code the way that you want it to be. It’s almost as if there’s a layer behind the code, which is the actual code. “You’re not actually writing the code,” Temkin adds. “You’re writing pleas to create that code, and you have to ask nicely. For example, if you call Zeus father of the gods, you can’t call him that again immediately because he doesn’t think you’re trying very hard.” “And then of course, to end a block of code, you have to call on Hades to collect the souls of all the unused variables. And so on,” Temkin says. The History of Esoteric Programming Languages Temkin continues a long-running tradition: esoteric languages date back to the early days of computing, with examples such as INTERCAL (1972), which had cryptic syntax, meaning coders often needed to plead with the compiler to run it. The scene gained momentum in 1993, with Wouter van Oortmerssen’s FALSE, in which most syntax maps to a single character. Despite this, FALSE is a Turing-complete language that allows creating programs as complex as any contemporary programming language. Its syntactical restrictions meant the compiler (which translates the syntax to machine-readable instructions) is only 1 kilobyte, compared to C++ compilers, which were generally hundreds of kilobytes. Exploring further, Chris Pressey wondered why code always had to be written from left to right and created Befunge in 1993. “It took the idea of the single-character commands and said if you’re going to have commands that are only one letter, why do we need to read it left to right?” says Temkin. “Why can’t we have code move a little bit to the right, then turn up, and then go off the page and come up off the bottom and so on?“ So Pressey decided to create a language that would be the most difficult language to build a compiler for,” Temkin continues. “I believe that was the original idea, allowing the code to turn in different directions and flow across the space.” Much of the mid-90s trend coincided with the rise of shareware, the demo scene, and the nascent days of the Internet, when it was necessary to program everything to be as small as possible to share it. “There’s definitely a lot of crossover between these things because they involve this kind of artistry, but also a kind of technical wizardry in showing, ‘Look how much I can do with this really minimal program,’” Temkin says. “What really interested me in esoteric languages specifically is the way that it’s community-based,” Temkin says. “If you make a language, it’s an invitation for other people to use the language. And when you make a language and somebody else shows you what’s possible to do with your language or discovers something new about it that you couldn’t have foreseen on your own.” One of Temkin’s esoteric languages uses a cuneiform script.Daniel Temkin You can play with many of Daniel’s languages on his website, as well as the Esoteric Languages Wiki, which raises the question: In the modern connected age, how does one create a shareable esoteric language? “It’s something that I’ve changed my attitude about over the years,” says Temkin. “Early on, I thought I had to write a serious compiler for my language. But now I think what’s really important is that people across different platforms and spaces can use it. So in general, I try to write everything in JavaScript when I can and have it run in the browser. If I don’t, then I tend to stick with Python as it has the largest user base. But I do get a little bored with those two languages.” “I realize there’s a certain irony there,” Temkin adds.

Belting your favorite song over prerecorded music into a microphone in front of friends and strangers at karaoke is a popular way for people around the world to destress after work or celebrate a friend’s birthday. The idea for the karaoke machine didn’t come from a singer or a large entertainment company but from Nichiden Kogyo, a small electronics assembly company in Tokyo. The company’s founder, Shigeichi Negishi, was singing to himself at work one day in 1967 when an employee jokingly told him he was out of tune. Figuring that singing along to music would help him stay on pitch, Negishi began thinking about how to make that possible. He had the idea to turn one of the 8-track tape decks his company manufactured into what is now known as the karaoke machine. Later that year, he built what would become the first such machine, which he called the Music Box. The 30-centimeter cube housed an 8-track player for four tapes of instrumental recordings and included a microphone to sing into. He sold his machine in 1967 to a Japanese trading company, which then sold it to restaurants, bars, and hotel banquet halls, where they used it as entertainment. The machine was coined karaoke in the 1970s to describe the act of singing along to prerecorded music. The term is a combination of two Japanese words: kara, meaning empty, and okesutora, meaning orchestra. In a few years, dedicated establishments known as karaoke bars began to open across Japan. Today the country has more than 8,000, according to Statista. The karaoke machine has been commemorated as an IEEE Milestone. The dedication ceremony was held in June in the area that houses karaoke booths connected to the Shinagawa Prince Hotel in Tokyo. Negishi’s family attended the event along with IEEE leaders. Negishi died last year at the age of 100. He was grateful that people enjoy karaoke around the world, his son, Akihiro Negishia, said at the ceremony, “though he didn’t imagine it to spread globally when he created it.” Accidentally inventing one of the world’s favorite pastimes Shigeichi Negishi grew up in Tokyo, where his mother ran a tobacco store and his father oversaw regional elections as a government official. After earning a bachelor’s degree in economics from Hosei University in Tokyo, he was drafted into the Imperial Japanese Army during World War II. He became a prisoner of war and spent two years in Singapore before being released in 1947. He returned to Tokyo and sold cameras for electrical parts manufacturer Olympus Corp. In 1956 he started Nichiden Kogyo, which manufactured and assembled portable radios for the home and car, according to the Engineering and Technology History Wiki entry about the karaoke machine. Negishi would start each morning singing along to the “Pop Songs Without Lyrics” radio show, according to a Forbes article. He typically didn’t sing in the office, but one fateful day he did. Negishi was inspired to engineer one of the 8-track tape decks his company manufactured into what is now known as the karaoke machine An 8-track tape deck can play and record audio using magnetic tape cartridges. Nichiden Kogyo’s Music Box was a 30-centimeter cube with slots to insert four 8-track tapes on the top panel, with control buttons to play, stop, or skip to the next song. Inside each 13-centimeter-long rectangular 8-track cartridge is a loop of almost 1 cm-wide magnetic tape that is coiled around a circular reel, as explained in an EverPresent blog post on the technology. A small motor inside each cartridge pulls the tape across an audio head inside the player, which reads the magnetic patterns and translates them into sound. Each tape had a metal sensing strip that notified a solenoid coil located in the player when a song had ended or if a person pressed the button to switch to the next song, according to an Autodesk Instribules blog post. The coil created a magnetic field when electricity passed through it—which rotated the spindle on which the audio head was mounted to move to the next track on the tape. Each tape could hold about eight songs. Negishi added a microphone amplifier to the player’s top panel, as well as a mixing circuit. The user could adjust the volume of the music and the microphone. He also recorded 20 of his favorite songs onto the tapes and printed out the lyrics on cardstock. He tested the machine by singing a popular ballad, “Mujo no Yume” (“The Heartless Dream”). “It works! That’s all I was thinking,” Negishi told reporter Matt Alt years later, when asked what his thoughts were the first time he tested the Music Box. Alt wrote Pure Invention: How Japan Made the Modern World. In 1969 engineers at Tokyo-based trading company Kokusai Shohin added a coin acceptor to the machine, renaming the Music Box the Sparko Box.Dr. Tomohiro Hase The fees to file a patent were too expensive, according to the ETHW entry, so in 1967 Negishi sold the rights to the machine to Mitsuyoshi Hamasu, a salesman at Kokusai Shohin. The Tokyo-based trading company began selling and leasing the machines by the end of the year. In 1969 engineers at Kokusai Shohin added a coin acceptor to the machine. The company renamed the Music Box the Sparko Box. In six years, about 8,000 units were sold, Hamasu said in an interview about the rise of karaoke. Karaoke became so popular that in the 1980s, venues and bars specializing in soundproofed rooms known as karaoke boxes emerged. Groups could rent the rooms by the hour. Negishi’s family owns the first Music Box he made. It still works. The Milestone plaque recognizing the karaoke machine is on display in front of the former headquarters of Nichiden Kogyo, which Negishi turned into a tobacco shop after he retired. The shop is now owned by his daughter. The plaque reads: “The first karaoke machine was created in 1967 by mixing live vocals with prerecorded accompaniment for public entertainment, leading to its worldwide popularity. Created by Shigeichi Negishi of Nichiden Kogyo, and originally called Music Box (later Sparko Box), it included a mixer, microphone, and 8-track tape player, with a coin payment system to charge the singer. An early operational machine has been displayed at the original company site in Tokyo.” Administered by the IEEE History Center and supported by donors, the Milestone program recognizes outstanding technical developments around the world. The IEEE Tokyo Section sponsored the nomination.

Millions of people worldwide have reason to be thankful that Swedish engineer Rune Elmqvist decided not to practice medicine. Although qualified as a doctor, he chose to invent medical equipment instead. In 1949, while working at Elema-Schonander (later Siemens-Elema), in Stockholm, he applied for a patent for the Mingograph, the first inkjet printer. Its movable nozzle deposited an electrostatically controlled jet of ink droplets on a spool of paper. Rune Elmqvist qualified to be a physician, but he devoted his career to developing medical equipment, like this galvanometer.Håkan Elmqvist/Wikipedia Elmqvist demonstrated the Mingograph at the First International Congress of Cardiology in Paris in 1950. It could record physiological signals from a patient’s electrocardiogram or electroencephalogram in real time, aiding doctors in diagnosing heart and brain conditions. Eight years later, he worked with cardiac surgeon Åke Senning to develop the first fully implantable pacemaker. So whether you’re running documents through an inkjet printer or living your best life due to a pacemaker, give a nod of appreciation to the inventive Dr. Elmqvist. The world’s first inkjet printer Rune Elmqvist was an inquisitive person. While still a student, he invented a specialized potentiometer to measure pH and a portable multichannel electrocardiograph. In 1940, he became head of development at the Swedish medical electronics company Elema-Schonander. Before the Mingograph, electrocardiograph machines relied on a writing stylus to trace the waveform on a moving roll of paper. But friction between the stylus and the paper prevented small changes in the electrical signal from being accurately recorded. Elmqvist’s initial design was a modified oscillograph. Traditionally, an oscillograph used a mirror to reflect a beam of light (converted from the electrical signal) onto photographic film or paper. Elmqvist swapped out the mirror for a small, moveable glass nozzle that continuously sprayed a thin stream of liquid onto a spool of paper. The electrical signal electrostatically controlled the jet. The Mingograph was originally used to record electrocardiograms of heart patients. It soon found use in many other fields.Siemens Healthineers Historical Institute By eliminating the friction of a stylus, the Mingograph (which the company marketed as the Mingograf) was able to record more detailed changes of the heartbeat. The machine had three paper-feed speeds: 10, 25, and 50 millimeters per second. The speed could be preset or changed while in operation. RELATED: The Inventions That Made Heart Disease Less Deadly An analog input jack on the Mingograph could be used to take measurements from other instruments. Researchers in disciplines far afield from medicine took advantage of this input to record pressure or sound. Phoneticians used it to examine the acoustic aspects of speech, and zoologists used it to record birdsongs. Throughout the second half of the 20th century, scientists cited the Mingograph in their research papers as an instrument for their experiments. Today, the Mingograph isn’t that widely known, but the underlying technology, inkjet printing, is ubiquitous. Inkjets dominate the home printer market, and specialized printers print DNA microarrays in labs for genomics research, create electrical traces for printed circuit boards, and much more, as Phillip W. Barth and Leslie A. Field describe in their 2024 IEEE Spectrum article “Inkjets Are for More Than Just Printing.” The world’s first implantable pacemaker Despite the influence of the Mingograph on the evolution of printing, it is arguably not Elmqvist’s most important innovation. The Mingograph helped doctors diagnose heart conditions, but it couldn’t save a patient’s life by itself. One of Elmqvist’s other inventions could and did: the first fully implantable, rechargeable pacemaker. The first implantable pacemaker [left] from 1958 had batteries that needed to be recharged once a week. The 1983 pacemaker [right] was programmable, and its batteries lasted several years.Siemens Healthineers Historical Institute Like many stories in the history of technology, this one was pushed into fruition at the urging of a woman, in this case Else-Marie Larsson. Else-Marie’s 43-year-old husband, Arne, suffered from scarring of his heart tissue due to a viral infection. His heart beat so slowly that he constantly lost consciousness, a condition known as Stokes-Adams syndrome. Else-Marie refused to accept his death sentence and searched for an alternative. After reading a newspaper article about an experimental implantable pacemaker being developed by Elmqvist and Senning at the Karolinska Hospital in Stockholm, she decided that her husband would be the perfect candidate to test it out, even though it had been tried only on animals up until that point. External pacemakers—that is, devices outside the body that regulated the heart beat by applying electricity—already existed, but they were heavy, bulky, and uncomfortable. One early model plugged directly into a wall socket, so the user risked electric shock. By comparison, Elmqvist’s pacemaker was small enough to be implanted in the body and posed no shock risk. Fully encased in an epoxy resin, the disk-shaped device had a diameter of 55 mm and a thickness of 16 mm—the dimensions of the Kiwi Shoe Polish tin in which Elmqvist molded the first prototypes. It used silicon transistors to pace a pulse with an amplitude of 2 volts and duration of 1.5 milliseconds, at a rate of 70 to 80 beats per minute (the average adult heart rate). The pacemaker ran on two rechargeable 60-milliampere-hour nickel-cadmium batteries arranged in series. A silicon diode connected the batteries to a coil antenna. A 150-kilohertz radio loop antenna outside the body charged the batteries inductively through the skin. The charge lasted about a week, but it took 12 hours to recharge. Imagine having to stay put that long. In 1958, over 30 years before this photo, Arne Larsson [right] received the first implantable pacemaker, developed by Rune Elmqvist [left] at Siemens-Elema. Åke Senning [center] performed the surgery.Sjöberg Bildbyrå/ullstein bild/Getty Images Else-Marie’s persuasion and persistence pushed Elmqvist and Senning to move from animal tests to human trials, with Arne as their first case study. During a secret operation on 8 October 1958, Senning placed the pacemaker in Arne’s abdomen wall with two leads implanted in the myocardium, a layer of muscle in the wall of the heart. The device lasted only a few hours. But its replacement, which happened to be the only spare at the time, worked perfectly for six weeks and then off and on for several more years. Arne Larsson lived another 43 years after his first pacemaker was implanted. Shown here are five of the pacemakers he received. Sjöberg Bildbyrå/ullstein bild/Getty Images Arne Larsson clearly was happy with the improvement the pacemaker made to his quality of life because he endured 25 more operations over his lifetime to replace each failing pacemaker with a new, improved iteration. He managed to outlive both Elmqvist and Senning, finally dying at the age of 86 on 28 December 2001. Thanks to the technological intervention of his numerous pacemakers, his heart never gave out. His cause of death was skin cancer. Today, more than a million people worldwide have pacemakers implanted each year, and an implanted device can last up to 15 years before needing to be replaced. (Some pacemakers in the 1980s used nuclear batteries, which could last even longer, but the radioactive material was problematic. See “The Unlikely Revival of Nuclear Batteries.”) Additionally, some pacemakers also incorporate a defibrillator to shock the heart back to a normal rhythm when it gets too far out of sync. This lifesaving device certainly has come a long way from its humble start in a shoe polish tin. Rune Elmqvist’s legacy Whenever I start researching the object of the month for Past Forward, I never know where the story will take me or how it might hit home. My dad lived with congestive heart failure for more than two decades and absolutely loved his pacemaker. He had a great relationship with his technician, Francois, and they worked together to fine-tune the device and maximize its benefits. And just like Arne Larsson, my dad died from an unrelated cause. An engineer to the core, he would have delighted in learning about the history of this fantastic invention. And he probably would have been tickled by the fact that the same person also invented the inkjet printer. My dad was not a fan of inkjets, but I’m sure he would have greatly admired Rune Elmqvist, who saw problems that needed solving and came up with elegantly engineered solutions. Part of a continuing series looking at historical artifacts that embrace the boundless potential of technology. An abridged version of this article appears in the September 2025 print issue. References There is frustratingly little documented information about the Mingograph’s origin story or functionality other than its patent. I pieced together how it worked by reading the methodology sections of various scientific papers, such as Alf Nachemson’s 1960 article in Acta Orthopaedica Scandinavica, “Lumbar Intradiscal Pressure: Experimental Studies on Post-mortem Material”; Ingemar Hjorth’s 1970 article in the Journal of Theoretical Biology, “A Comment on Graphic Displays of Bird Sounds and Analyses With a New Device, the Melograph Mona”; and Paroo Nihalani’s 1975 article in Phonetica, “Velopharyngeal Opening in the Formation of Voiced Stops in Sindhi.” Such sources reveal how this early inkjet printer moved from cardiology into other fields. Descriptions of Elmqvist’s pacemaker were much easier to find, with Mark Nicholls’s 2007 profile “Pioneers of Cardiology: Rune Elmqvist, M.D.,” in Circulation: Journal of the American Heart Association, being the main source. Siemens also pays tribute to the pacemaker on its website; see, for example, “A Lifesaver in a Plastic Cup.”

The last time I used a dial-up modem came sometime around 2001. Within just a few years, dial-up had exited my life, never to return. I haven’t even had a telephone line in my house for most of my adult life. But I still feel a strong tinge of sadness to know that AOL is finally retiring the ol’ hobbyhorse. At the end of September, it’s gone. The timeline is almost on-the-nose fitting: The widespread access to the Internet AOL’s service brought in the 1990s is associated with a digital phenomenon called the Eternal September. Before AOL allowed broad access to Usenet—a precursor to today’s online discussion forums—most new users appeared each September, when new college students frequently joined the platform. Thanks to AOL, they began showing up daily starting around September 1993. The fact that AOL’s dial-up is still active in the first place highlights a truism of technology: Sometimes, the important stuff sticks around well after it’s obsolete. Why AOL is ditching dial-up now It’s no surprise that dial-up has lingered for close to a quarter-century. Despite not having needed a dial-up modem myself since the summer of 2001, I was once so passionate about dial-up that I begged to get a modem for my 13th birthday. Modems are hard to shake, and not just because we fondly remember waiting so long for them to do their thing. Originally, the telephone modem was a hack. It was pushed into public consciousness partly by Deaf users who worked around the phone industry’s monopolistic regulations to develop the teletypewriter, a system to communicate over phone lines via text. Along the way, the community invented technologies like the acoustic coupler. To make that hack function, modems had to do multiple conversions in real time—from data to audio and back again, in two directions. As I put it in a piece that compared the modem to the telegraph: The modem, at least in its telephone-based forms, represents a dance between sound and data. By translating information into an aural signal, then into current, then back into an aural signal, then back into data once again, the modulation and demodulation going on is very similar to the process used with the original telegraph, albeit done manually. Modems like this one from U.S. Robotics work by converting data to audio and back again. Jphill19/Wikimedia Commons With telegraphs, the information was input by a person, translated into electric pulses, and received by another person. Modems work the same way, just without human translators. The result of all this back and forth was that modems had to give up a hell of a lot of speed to make this all work. The need to connect over a medium built for audio meant that data was at risk of getting lost over the line. (This is why error correction was an essential part of the modem’s evolution; often data needed to be shared more than once to ensure it got through. Without error correction, dial-up modems would be even slower.) Remember that sound? It marked many users’ first experience getting online.AdventuresinHD/YouTube Telephone lines were a hugely inefficient system for data because they were built for voice and heavily compressed audio. Voices are still clear and recognizable after being compressed, but audio compression can wreak havoc on data connections. Plus, there was the problem of line access. With a call, you could not easily share a connection. That meant you couldn’t make phone calls while using dial-up, leading to some homes getting a second line. And at the Internet Service Provider level, having multiple lines got very complex, very fast. The phone industry knew this, but its initial solution, ISDN, did not take off among mainstream consumers. (A later one, DSL, had better uptake, and is likely one of the few Internet options rural users currently have.) In some areas of the United States, dial-up remains the best option—the result of decades of poor investment in Internet infrastructure. So the industry moved to other solutions to get consumers Internet—coaxial cable, which was already widespread because of cable TV, and fiber, which wasn’t. The problem is, coax never reached quite as far as telephone wires did, in part because cable television wasn’t technically a utility in the way electricity or water were. In recent years, many attempts have been made to classify Internet access as a public utility, though the most recent one was struck down by an appeals court earlier this year. The public utility regulation is important. The telephone had struggled to reach rural communities in the 1930s, and only did so after a series of regulations, including one that led to the creation of the Federal Communications Commission, were put into effect. So too did electricity, which needed a dedicated law to expand its reach. But the reach of broadband is frustratingly incomplete, as highlighted by the fact that many areas of the country are not properly covered by cellular signals. And getting new wires hung can be an immensely difficult task, in part because companies that sell fiber, like Verizon and Google, often stop investing due to the high costs. (Though, to Google’s credit, it started expanding again in 2022 after a six-year rollback.) So, in some areas of the United States, dial-up remains the best option—the result of decades of poor investment in Internet infrastructure. This, for years, has propped up companies like AOL, which has evolved numerous times since it foolishly merged with Time Warner a quarter-century ago. The first PC-based client called America Online appeared on the graphical operating system GeoWorks. This screenshot shows the DOS AOL client that was distributed with GeoWorks 2.01.Ernie Smith But AOL is not the company it was. After multiple acquisitions and spin-outs, it is now a mere subsidiary of Yahoo, and it long ago transitioned into a Web-first property. Oh, it still has subscriptions, but they’re effectively fancy analogues for unnecessary security software. And their email client, while having been defeated by the likes of Gmail years ago, still has its fans. When I posted the AOL news on social media, about 90 percent of the responses were jokes or genuine notes of respect. But there was a small contingent, maybe 5 percent, that talked about how much this was going to screw over far-flung communities. I don’t think it’s AOL’s responsibility to keep this model going forever. Instead, it looks like the job is going to fall to two companies: Microsoft, whose MSN Dial-Up Internet Access costs US $179.95 per year, and the company United Online, which still operates the longtime dial-up players Juno and NetZero. Satellite Internet is also an option, with older services like HughesNet and newer ones like Starlink picking up the slack. It’s not AOL’s fault. But AOL is the face of this failing. AOL dropping dial-up is part of a long fade-out As technologies go, the dial-up modem has not lasted quite as long as the telegram, which has been active in one form or another for 181 years. But the modem, which was first used in 1958 as part of an air-defense system, has stuck around for a good 67 years. That makes it one of the oldest pieces of computer-related technology still in modern use. To give you an idea of how old that is: 1958 is also the year that the integrated circuit, an essential building block of any modern computer, was invented. The disk platter, which became the modern hard drive, was invented a year earlier. The floppy disk came a decade later. (It should be noted that the modem itself is not dying—your smartphone has one—but the connection your landline has to your modem, the really loud one, has seen better days.) The news that AOL is dropping its service might be seen as the end of the line for dial-up, but the story of the telegram hints that this may not be the case. In 2006, much hay was made about Western Union sending its final telegram. But Western Union was never the only company sending telegrams, and another company picked up the business. You can still send a telegram via International Telegram in 2025. (It’s not cheap: A single message, sent the same day, is $34, plus 75 cents per word.) In many ways, AOL dropping the service is a sign that this already niche use case is going to get more niche. But niche use cases have a way of staying relevant, given the right audience. It’s sort of like why doctors continue to use pagers. As a Planet Money episode from two years ago noted, the additional friction of using pagers worked well with the way doctors functioned, because it ensured that they knew the messages they were getting didn’t compete with anything else. Dial-up is likely never going to totally die, unless the landline phone system itself gets knocked offline, which AT&T has admittedly been itching to do. It remains one of the cheapest options to get online, outside of drinking a single coffee at a Panera and logging onto the wifi. But AOL? While dial-up may have been the company’s primary business earlier in its life, it hasn’t really been its focus in quite a long time. AOL is now a highly diversified company, whose primary focus over the past 15 years has been advertising. It still sells subscriptions, but those subscriptions are about to lose their most important legacy feature. AOL is simply too weak to support the next generation of Internet service themselves. Their inroad to broadband was supposed to be Time Warner Cable; that didn’t work out, so they pivoted to something else, but kept around the legacy business while it was still profitable. It’s likely that emerging technologies, like Microsoft’s Airband Initiative, which relies on distributing broadband over unused “white spaces” on the television dial, stand a better shot. 5G connectivity will also likely improve over time (T-Mobile already promotes its 5G home Internet as a rural option), and perhaps more satellite-based options will emerge. Technologies don’t die. They just slowly become so irrelevant that they might as well be dead. The monoculture of the AOL login experience When I posted the announcement, hidden in an obscure link on the AOL website sent to me by a colleague, it immediately went viral on Bluesky and Mastodon. That meant I got to see a lot of people react to this news in real time. Most had the same comment: I didn’t even know it was still around. Others made modem jokes, or talked about AOL’s famously terrible customer service. What was interesting was that most people said roughly the same thing about the service. That is not the case with most online experiences, which usually reflect myriad points of views. I think it speaks to the fact that while the Internet was the ultimate monoculture killer, the experience of getting online for the first time was largely monocultural. Usually, it started with a modem connecting to a phone number and dropping us into a single familiar place. We have lost a lot of Internet Service Providers over the years. Few spark the passion and memories of America Online, a network that somehow beat out more innovative and more established players to become the onramp to the Information Superhighway, for all the good and bad that represents. AOL must be embarrassed of that history. It barely even announced its closure.

CT scanning, streaming videos, and sending images over the Internet wouldn’t be possible without the Fast Fourier transform. Commonly known as FFT, the computer algorithm designed by researchers at Princeton University and IBM is found in just about every electronic device, according to an entry in the Engineering and Technology History Wiki. Demonstrated for the first time in 1964 by IEEE Fellows John Tukey and James W. Cooley, the algorithm breaks down a signal—a series of values over time—and converts it into frequencies. FFT was 100 times faster than the existing discrete Fourier transform. The DFT also requires more memory than the FFT because it saves intermediate results while processing. The FFT has become an important tool for manipulating and analyzing signals in many areas including audio processing, telecommunications, digital broadcasting, and image analysis. It helps filter, compress, eliminate noise from, and otherwise modify signals. The 60-year-old ubiquitous computer code also has applications in today’s cutting-edge technologies such as AI, quantum computing, self-driving cars, and 5G communication systems. The FFT was commemorated with an IEEE Milestone during a ceremony held in May at Princeton University. “The Cooley-Tukey algorithm significantly accelerated the calculation of DFTs,” 2024 IEEE President Tom Coughlin said at the ceremony. “Prior methods required significantly more computations, making FFT a revolutionary breakthrough. By leveraging algebraic properties and periodicities, the FFT reduced the number of the operations, making it particularly and practically feasible for everyday tasks, replacing the less efficient analog methods.” A new mathematical tool In 1963 Tukey, a professor of mathematics and statistics at Princeton, participated in a meeting of U.S. President John F. Kennedy’s Science Advisory Committee to discuss ways to detect underground nuclear tests, according to the ETHW entry. Also attending that meeting was Richard Garwin, a physicist and engineer at IBM who played a key role in designing the first hydrogen bomb. He died in May. Read about his fascinating life in this month’s In Memoriam. Tukey told Garwin he was working on speeding up the computation of an existing method—the Fourier transform—thinking it might help with the detection. His algorithm mathematically converted a signal from its original domain, such as time or space, to a frequency domain. Garwin recognized its potential and asked IBM to select a mathematical analyst to collaborate with Tukey. That person was Cooley, a research staff member working on numerical analysis and computation projects. If the Fourier transform could be made faster, Garwin said, seismometers could be planted in the ground in countries surrounding the Soviet Union to detect nuclear explosions from atomic bomb tests, because the Soviets wouldn’t allow on-site tests, according to Cooley’s oral history in the Engineering and Technology History Wiki. A seismometer measures ground vibrations, which are converted into electrical signals and recorded as seismograms. To design sensors for underground nuclear tests, however, “you would have to process all the seismic signals, and a large part of the processing could be done by Fourier transforms,” Cooley said in his oral history. But “the computing power at the time was not enough to process all of the signals you’d need to do this.” The FFT could calculate a seismic sensor’s frequency and produce images, IEEE Life Fellow Harold S. Stone said at the Milestone event. He is an image processing researcher and Fellow emeritus at the NEC Laboratories America, in Princeton, and a former IBM researcher. Tukey and Cooley led the team that wrote the computer code that demonstrated the FFT’s power. “The demonstration of the Coley-Tukey algorithm showed that it was 100 times faster,” Stone said. “It was so fast that it could keep up with the seismic data.” Sensors using the algorithm were planted, and they detected nuclear explosions within a 15-kilometer radius from where they were detonated, according to the ETHW entry. “By leveraging algebraic properties and periodicities, the FFT reduced the number of the operations, making it particularly and practically feasible for everyday tasks, replacing the less efficient analog methods.” —2024 IEEE President Tom Coughlin In 1965 Cooley and Tukey published “An Algorithm for the Machine Calculation of Complex Fourier Series,” describing the FFT process. The seminal paper spurred development of digital signal processing technologies. For his work, Tukey was awarded a U.S. National Medal of Science in 1973. He also received the 1982 IEEE Medal of Honor for “contributions to the spectral analysis of random processes and the fast Fourier transform algorithm.” Cooley, who received the 2002 IEEE Kilby Signal Processing Medal for pioneering the FFT, was a leading figure in the field of digital signal processing. Through his involvement with the IEEE Digital Signal Processing Committee (today known as the IEEE Signal Processing Society), he helped establish terminology and suggested research directions. Although not one of the inventors, Garwin is credited with recognizing that the algorithm had wider applications, especially in scientific and engineering fields. “In today’s lingo, Garwin helped the FFT ‘go viral’ by getting Cooley and Tukey together,” Stone said. “Garwin and Tukey sought better information to forestall and prevent wars,” added Frank Anscombe, Tukey’s nephew. “The Cooley-Tukey FFT swiftly advanced this cause by giving a practical, simplifying solution for wavy data. Thanks to the FFT, a technological rubicon began to be crossed: analog-to-digital machines.” A spirit of collaboration between academia and industry Like so many innovations, the FFT came out of a collaboration between industry and academia, and it should be recognized for that, IEEE Fellow Andrea Goldsmith said at the ceremony. She explained that she regularly works with FFT in her research projects. At the time of the event, she was Princeton’s dean of engineering and applied sciences. This month she started her new position as president of Stony Brook University, in New York. “Taking the ideas we have from basic research in our university labs, talking to people in industry, and understanding how the research problems we work on can benefit industry either tomorrow or in five years or 20 years from now, is incredibly important,” she said. “Some people think of engineering as boring and dry and something that only nerds do, but there is such beauty and creativity in a lot of the innovations that we have developed, and I think the FFT is a perfect example of that.” The FFT joins more than 270 other IEEE Milestones. They are more than a marker of achievement, said IEEE Life Senior Member Bala S. Prasanna, director of IEEE Region 1. “They are a testament to human ingenuity, perseverance, and the spirit of collaboration,” Prasanna said. “These Milestones were more than just breakthroughs; they became catalysts for innovation, enabling progress in ways once thought impossible. Each one ensures that the story behind these innovations is preserved, not just as history but as inspiration for future generations.” Another ceremony was held on 11 June at the IBM Watson Research Center. Milestone plaques recognizing the FFT are on display in the lobby of Princeton’s School of Engineering and Applied Science and in the main lobby at the entrance of the IBM research center. They read: “In 1964 a computer program implementing a highly efficient Fourier analysis algorithm was demonstrated at IBM Research. Jointly developed by Princeton University and IBM collaborators, the Cooley-Tukey technique calculated discrete Fourier transforms orders of magnitude faster than had been previously demonstrated. Known as the Fast Fourier Transform (FFT), its speed impacted numerous applications including computerized tomography, audio and video compression, signal processing, and real-time data streaming.” Administered by the IEEE History Center and supported by donors, the Milestone program recognizes outstanding technical developments around the world. The IEEE Princeton Central Jersey Section sponsored the nomination.