Linkfest #44: Hallucinopedia, Salmon on Cocaine, and How Phil Collins Accidentally Invented Gated Reverb
Hello there!
Time for "the opposite of doomscrolling” — my latest “Linkfest”, in which I scroll through the endless Borgesian microfiche of the Internet, looking for the finest items of science, culture and technology.
These would make a pretty fun gift. It’d be interesting if someone could build a tool that would let you scan the cover of any album, generate a pixelated version, then tell you the quantity of bricks you’d need and the pattern in which to place them …
“Uncrossy” is a really fun word game. You drag a word sideways (or up and down) so that it forms a new one, at which point it vanishes. The goal is to get all the words to vanish.
There are a bunch of tricky bits; you can’t move a word that’s moored at both ends, for example. So you have to sort of play around and gradually figure out what moves are possible. This experimental quality means you rarely get stuck, which gives the game a nice casual feel.
I've been a puzzle designer for over 2 decades, making games like Cogs, which launched in 2009. I love games with simple rules and complex behavior that make you say, "I should have thought of that!"
The tricky thing about Uncrossy is not the rules of play but how the puzzles are designed. I basically need to design puzzles in reverse and have built a set of tools to help me with that.
3) 🥁 How Phil Collins accidentally invented “gated reverb”, the sound of 80s drumming
When Phil Collins was in the studio drumming on Peter Gabriel’s third album, the recording engineer inadvertently left the talkback microphone on. The talkback microphone had heavy compression, and it accidentally produced a fascinating new type of reverb.
Normally, the reverb on a drum hit is intense at first, then slowly fades away. But in the Phil Collins drum session, the accidental reverb behaved differently. It had a nice, loud, booming reverb for a moment — then the reverb abruptly stopped. This created a very cool new type of drum sound. It was boomy and huge, but wasn’t messy, because the reverb for each drum hit ended before the next drum hit.
They dubbed it “gated reverb”: The reverb starts going, but you quickly slam the gate down and stop it. And it’s a very weird, artificial sound! Reverb doesn’t behave like this in the real world.
But Phil Collins was mesmerized by it. He began using it on his own music, and in 1981 he famously used it for the powerhouse drumming of “In The Air Tonight”. That song was such a monster hit that artists quickly began using gated reverb, and it became the dominant sound of the 80s.
Needless to say, Prince loved the sound of gated reverb. From about 1982 to 1987, the technique was all over his records! He was also a fan of sampled drums, and famously fed a Linn LM-1 drum machine into the RMX16 reverb unit to create the quintessential ’80s drum sound heard on some of his biggest albums.
At this point, gated reverb was no longer an accident at all—it had become the defining sound of an era. Some of the most popular examples of this drum sound can be heard on “Hounds of Love” by Kate Bush, “Jack and Diane” by John Cougar Mellencamp, “Dance on Your Knees” by Hall and Oates, and “A View to Kill” by Duran Duran.
By the 90s, people were sick of it, and drumming became “dry”, with no reverb. But fashion is cyclical, so heyo, by the 2010s …
Projects like “When I Needed You” by Carly Rae Jepsen, Melodrama by Lorde, and the appropriately titled 1989 by Taylor Swift all prominently feature the ’80s drum sound which Hugh Padgham and Phil Collins accidentally created back in 1979.
The study took a bunch of two-day-old infants and put EEG sensors on their skulls while they slept. The scientists played two types of music for the babies — Bach songs, and “scrambled” songs with random notes at random timing.
EEG measurements can detect when someone’s brain registers surprise. They found that the sleeping infants registered surprise when the rhythmic pattern of a Bach song changed. Basically, this indicated that the infants were noticing the rhythms of the songs — and paying such close attention that they could detect when Bach changed things.
What’s more, they didn’t have these surprise reactions to the jumbled songs. Those songs didn’t have rhythmic structure. So this helps prove that the babies really were responding to rhythm and changes in rhythm.
Bianco said the findings suggested the human brain was biologically tuned to make predictions when listening to music, especially about rhythm.
“Importantly, these predictions go beyond simply anticipating a regular interval: they involve detecting patterns in the music and learning how those patterns unfold over time,” she added.
Bianco said such abilities in newborns probably had their roots in very basic biological and sensory experiences. “Before birth, the foetal environment is dominated by regular rhythms, such as the mother’s heartbeat and the repeated motion associated with her walking,” she said, noting such rhythms may provide the brain with an early sense of timing and predictability.
Another interesting finding: Infants didn’t react to changes in the melody. So infants can predict rhythm, but haven’t yet learned how to predict melody.
“After a few prototypes to test the material, I discovered that carbon based black pasta worked quite well,” the pasta-making photographer says. “Even so, I had several issues with light leaks. Some of the first photographs came out very gray and lacking contrast. With the final prototype, the one shown in the pictures, I finally achieved some solid results.” [snip]
“I’ve always liked pinhole photography and have been fascinated by how people can make working cameras out of almost anything. Before this project, though, I had only used beer cans for solargraphy. In fact I usually do more street photography and I really enjoy shooting on film. I have a darkroom and I genuinely appreciate the often frustrating process of working in it. Like cooking, it requires patience and care,” Ambrogi says.
He calls it the “RaviHole”, lol. I really love how it gives everything a spooky resonance!
via Paride Ambrogi
Maybe I’ll make my own pastacam, take a selfie, and use it as the author photo for my next book.
6) 🦇 Poster showing the evolution of the Batman logo
The early evolution is slower than you'd expect. From 1939 to the early 1960s, the bat-symbol changed mostly in proportion — wings got wider, then narrower, then wider again. The head appeared, disappeared, grew prominent, shrank back. Artists tweaked wing points from five to seven to nine without much consistency. Printing technology was crude enough that fine details often vanished on the page anyway.
Then 1964 changed everything. The bat landed inside a bright yellow ellipse, and suddenly the logo had presence. The version refined in 1966 — with the wings curving outward to fill the oval — became the definitive Batman emblem for an entire generation. It held that position for over three decades.
DC's alternate universes — Elseworlds, the Dark Multiverse, one-shot specials — are where the bat-symbol gets truly strange. Batman: Holy Terror reimagines Bruce Wayne as a priest, and the emblem reflects it. Batman: Digital Justice #4, the first fully digital comic book ever made, carries its own distinct symbol. The 2017 Dark Nights: Metal event spawned an entire gallery of corrupted Batman variants — the Dawnbreaker, the Drowned, the Merciless, the Devastator — each with an emblem designed to feel wrong, like a bat-symbol from a universe where Batman lost.
7) 👻 Horror novel where no two copies are the same
Subcutanean is a coming-of-age horror novel that Aaron Reed published in 2020, which tells of a college senior who discovered a secret basement beneath his house.
The wild thing is that every copy of the novel is slightly different. Reed wrote software that let him slightly change the text on a word-by-word and passage-by-passage level every time he printed a copy for sale.
Don't expect entirely different novels. Each version of Subcutanean has the same number of chapters, for instance, and each chapter has essentially the same beats: there's no branching paths or radically alternate endings. You might think of each version as a different draft of the same book. As a data point, my master copy with all possible versions is about 100,000 words long: any particular rendering will contain about 62,000 words.
This is not to say all the changes are purely cosmetic, though. Each of the 17 chapters has at least one major moment that differs from version to version: some versions might include extra scenes or entirely different ones, or have pivotal moments that can play out in meaningfully different ways. Some of these alternates have a 50/50 chance of showing up: others are more rare, maybe appearing in only one book out of ten. All told this makes for several hundred thousand possible combinations of the major variants alone—with countless more once you factor in all the smaller alternates, of which there are generally at least one per paragraph, sometimes far more.
To write the novel, Reed created his own custom markup language — .quant — that let him write a passage while quickly denoting possible word/clause substitutions and their probabilities of being generated. His essay about the process is incredibly interesting. For example, this line can produce either the sentence “In the end, I almost forgave him” or “In the end, I forgave” — which hit in very different ways …
He can also assign relative probabilities of a word/clause occurring — he gives this example (which isn’t from the book itself) …
I really dig the fact that Reed didn’t try to write branching plots. I personally find branching plots kind of dull. I mean, it’s sufficiently hard for an author to come with a single unitary plot that’s truly satisfying; asking them to generate many (dozens! scores!) just kinda waters things down, for me. In contrast, authors regularly tinker intensively with sentences and passages, trying out different gambits of phrasing and emphasis — and they often have real trouble settling on their final one. So a software-based generator that captures that sort of literary noodling? Much more interesting, to me!
For example, comparing Seed 60001 with Seed 60025, we can see in Chapter 2 how a single word change—hangover/memory—alters what we know about Ryan. Similarly, our understanding of Niko changes if we know he has board game buddies instead of philosophy buddies.
Incredibly interesting stuff. I want to try writing some poetry using .quant now.
Disposable vapes are becoming a huge source of e-waste, because they contain batteries and microchips. At the “Vape Synth” project, the creators figured out a way to repurpose them — by turning them into a synthesizer.
Think of it like a digital ocarina. The Vape Synth repurposes the vaporizer's existing low-pressure sensor. By sucking wind through the sensor—maybe it's a reverse digital ocarina—you trigger an oscillator circuit and generate an audio signal. Pressing the buttons triggers different tones. [snip]
The people who made the Vape Synth know it sounds goofy. That’s the point.
“We started from a very silly place,” says Kari Love, one of Vape Synth’s creators. “We have to use the low pressure sensor. Which means to play it, you must suck.”
The Vape Synth’s tonal quality is a bit lacking. I found that it sort of sounds like a dying rabbit or swarm of flies hitting a bug zapper nonstop. Love says the team is already working on another version of the Vape Synth with a wider musical range that can also be used as a MIDI controller, and therefore able to trigger more pleasing sounds.
I’d love to try using it as a MIDI controller. I’m a harmonica player, and for years I played harmonica on a harmonica stand while busking on the streets of Toronto during college. Recently I discovered that a Swedish company makes the DM48X a MIDI device in the form of a harmonica, and whoooaaa I want to try this out! The idea of controlling a synth via a mouth-controlled device while playing guitar really intrigues me. Mind you, the DM48X is about $1,000 new, so … maybe I’ll wait on this one lol
9) 🎲 Native Americans played dice 18,000 years ago
via Robert Madden
For a long time, the earliest examples of dice — and games of chance — were from about 5,500 years ago in Mesopotamia.
But Robert Madden, a doctoral student at Colorado State University, just published the results of an exhaustive study he did of Native American artifacts — and he found examples of dice going back as far as 13,000 years.
The study says such activities, based on the concepts of chance, randomness and probability, were first developed in the present-day Southwestern U.S. — not in Mesopotamia or by other ancient cultures.
“We see it right in North America, beginning 12,000 years ago, people really starting to engage with some really complex kinds of intellectual concepts that aren’t grappled with in the Old World until many thousands of years later,” said Robert Madden, a doctoral student at Colorado State University, who is the study’s author. “These concepts end up being foundational to our modern scientific understanding, our modern economy.”
Archaeologists and historians had known for decades that Native Americans were huge fans of games of chance — there are written accounts dating back to the 1600s, and games were often played when far-flung groups were meeting and trading. Gambling was a quick way to bond …
The records, many from eastern North America, suggest the games were fast-paced, often played among unfamiliar groups.
“These tended to be very social affairs, very raucous affairs,” he said. “You tend to have a large crowd be gathered around the game and all kinds of side bets going on during the contest.”
As Madden notes, it’s impossible to know whether these early native groups were calculating the probabilities of the games …
Nonetheless, he said, his research offers evidence that Native Americans were doing complex counting and were likely to have been the first humans to contemplate concepts like the law of large numbers, a mathematics concept that describes how a random sample will trend toward an equal distribution over time.
“What we’re really looking at here is an intellectual accomplishment,” he said.
The original paper — “Probability in the Pleistocene: Origins and Antiquity of Native American Dice, Games of Chance, and Gambling” — is here for free.
An infinite, hallucinated encyclopedia. Every link leads to an entry that does not exist yet — until you click it, at which point an LLM pretends it has always existed and writes it for you, in the deadpan register of a 19th-century scholarly press.
It’s a witty way to showcase the drawbacks of LLMs, i.e., their propensity to offer confident-sounding nonsense when you ask them a question.
Though: As this thread at Hacker News discovered, Halupedia also showcases another bleak side-effect of the LLM-ification of our infosphere — which is that it’ll also happily fabricate truly awful content. So, be forewarned if you click through the list of generated articles! Most of them appear to be pretty silly, but there will likely be some rather septic entries, too.
My wife and I frequented the North Tunbridge General Store, and in early 2024, I approached Mike Gross, the store owner, about this project. I explained if he just gives me internet, I’ll provide the phone, pay for the service, and make sure the network is secure and that phone traffic is encrypted. And he was like, “I don’t know that I would want people messing around with the phone on my porch after hours.” But his wife, Lois, who’s a co-owner of the store, was like, “Oh, that would be great because people’s cars break down sometimes, and we get people asking to use the phone occasionally.” Also, around this time, there was a woman at a campground who had a dispute with her partner and ended up walking 10 or 15 miles to the store. She really could have used a phone. A couple of weeks later, I went to the store, and Mike was like, “So, I’ve been thinking about your phone project—let’s do it.” In March 2024, that first phone went in.
As he notes, the ergonomics of pay phones are amazing — at a glance, people know what they are and what they can/should do. And they’re built to last:
The biggest reason is, if you’re old enough to know what a pay phone is, and you see one of these old Western Electric and GTE pay phones, you instantly know that’s public infrastructure for you. You don’t have to ask for permission. You can just go up and use it. If it’s just a courtesy phone, it’s not always 100 percent clear what that’s for or if it’s okay to use. Also, pay phones are cool. There’s a lot of history behind them. And pay phones are rugged. They were built to withstand abuse and be outdoors for decades.
Pay phones truly are a marvel of industrial design. Frankly, those old phone handsets are one of the most gorgeous pieces of audio equipment, because of the way they cradled your head: One speaker perfectly positioned in front of your mouth, one earpiece right next to your ear. The handset felt like it was listening to you.
When people first started using mobile phones that were shaped like tiny blocks — the early Nokia phones, then the glass-slab-ification of the smartphone era — folks in public often shouted into the phones. It was incredibly annoying, but I think part of it was the design of the phones. They didn’t feel like they were listening to you. So it was like, well, I’m holding a stapler to my head — of course I’ve got to shout, if I’m going to be heard!
But payphones? Amazing design. Single-use devices FTW.
12) 🌐 Trust circle for proving your site was written by humans
Basically, it works like this: You put a little human.json file on your web site to verify that it was written by humans, not bots. You can also list other sites that you trust to be written by humans … you’re vouching for them. Other web site authors can similarly vouch for you.
For folks reading web sites, Dealmeida wrote a browser plugin that checks the human.json file site you’re on, and tells you if it claims to have been written by a human. You can decide to trust the site and click “trust”, and thereby build up a list of sites you believe are human-authored.
Now obviously, a website owner could lie, right? They could create their site with AI slop and then put a human.json file on and claim it was written by humans.
But Dealmeida’s theory is that the vouching will help disambiguate the AI sites from human sites. The plugin shows you if the site you’re on is vouched for by sites you trust. So there are positive incentives all around: Real human bloggers will want to vouch for other real human bloggers. They won’t vouch for AI-authored sites.
… if someone trusts you, they can expand the scope of that trust to the people that you're vouching for. People shouldn't blindly trust a human.json file, since anyone can publish one. But by trusting a few seed websites that they know are maintained by real humans, they can discover other real humans by following their vouches.
Sure, AI-authored sites could create their own networks and vouch for each other. But they’ll be in their own little septic pond. No real, human bloggers will vouch for them. It’s a bit like how Mastodon works; if someone sets up a server for horrible trolls, other Mastodon servers will “de-federate” from them, and they’ll be left in their own isolation.
This is obviously an early-stage idea. But this general concept, of devising social proofs? Disambiguating human creation from synthetic material? I would expect we’ll see more and more of such proposals. I’m super interested to see what people come up with!
13) 🛴 A final, sudden-death round of reading material
![A screenshot of a sentence reading: "In the end, I [almost] forgave him."](https://assets.buttondown.email/images/38974554-b4c1-4054-8d25-f145450f9058.png?w=960&fit=max)
![A screenshot of a sentence reading "Hey, [50>Charles|45>Chuck|5>C-Dog], how's it going?" I said.](https://assets.buttondown.email/images/c7b8fba8-514b-45c2-9b4b-20965d86267a.png?w=960&fit=max)
