After Rony Abovitz blew $3 billion running an ill-fated eyewear company, I found myself as a stop on his rebranding tour. He asked me to dream big and send a pitch.
Sharing so that the less-funded among us can enjoy it as entertainment, too.
From: Erik Gavriluk
Date: Mon, Jul 12, 2021 at 6:33 PM
Subject: La la la la – oh, what fun!
To: Rony Abovitz
Traditional hand tools set the standard for user experience. We use them to modify our environment and gather information about it. Being truly physics-based by virtue of existing in the non-virtual world, they provide feedback free from latency or quantization effects.
The tech industry continues to accelerate away from these principles. Even Apple suddenly found themselves unable to create a functional laptop keyboard. It took them four product generations to get back to baseline. Did they lose their way? Or did using their own tools lead them astray?
To explore, let’s go from Cupertino to Japan via Google, who credits Masaya Matsuura as a musician – complete with a photo of him in front of an ultra-rare analog synthesizer.
He’s better known as the digital inventor of the multi-billion dollar “press buttons in time to music” game genre, with multiple hit titles and at least one underground cultural marvel that recently got rediscovered and blasted all over TikTok.
But why did a 20 year old game with such primitive visuals suddenly explode in popularity all over the web?
Character design? Maybe… once you get past the pixels. Music and voice effects? Fantastic but arcane. Game design? Um, the whole thing started as a quick one-off for a car commercial.
It’s tempting to credit nostalgia but the people most excited about the game were born long after it was created.
I suggest instead that it’s yet another indicator of deep cultural awareness in the youngest generation. Unlike older people they don’t see the rabbit as retro so much as get a feeling that, whatever the heck it is, it probably shouldn’t look quite like that. They know something is going on, but more importantly they know something is missing. And they’re hungry for it.
In fewer than 30 lines of code I made this. And now young and old alike immediately get it – and feel it.
Now let’s travel from the Sony Playstation offices in Tokyo down to Shizuoka. In 1991, Mr. Suzuki bought all the 1930s tooling needed to build Hammond Organs. Today Hammond-Suzuki is the sole manufacturer of electronic music keyboards that have proper key travel. Everything else is awful and, just like the Apple keyboard mess, musicians complain about feel and, yes, even keys that play notes multiple times. Yet this music creation equipment costs thousands of dollars more than even Apple products. We have lost our way.
Above, the mechanism invented by what was then the Hammond Clock Company reveals a series of contact switches that connect nine busses to a crosspoint of electromagnetic pickups located near 91 precision gears ground at the clock factory. Each gear is locked to a capstan driven by a synchronous motor spinning at the 60 Hz electrical line frequency.
When pressing a key, the performer engages these 9 switches gradually, changing the attack and tone of the note. The mechanism also generates a distinctive “keyclick” sound and, unsurprisingly, no one is happy with the digital versions. Makes sense when you have a peek through the plexiglass (itself invented in the 1930s).
Later, different gear ratios were needed for 50 Hz countries. As a result, emulations of classic 1940s American Jazz equipment digitized by 1990s Eurodance software companies can sound a bit disconcerting to people who grew up listening to R&B records – regardless of where they grew up or what voltage comes out of their wall.
“…free from latency or quantization effects…”
Pianos are a relatively recent phenomenon – the organ precedes it by at least six centuries. At this point of the diatribe you might imagine how things worked out when, back in 1983, technical people invited themselves to put a layer between musicians and computers and create a Musical Instrument to Digital Interface.
The engineers behind MIDI oversimplified everything to “note on” and “note off.” Because digital. Being enthusiasts they couldn’t resist adding stuff like “after-touch” and other unnecessary abstractions. Yet they couldn’t come up with a bi-directional protocol. As a result, actual PANIC buttons were placed on instruments to stop commonplace cacophonies of beep-bloops from assaulting audiences when things fell out of sync.
A slow serial link was used, inspiring such technical innovations as “groove quantization” whereby a 1980s Atari would attempt to add “groove” back to your performance after losing the nuances during capture and playback due to latency.
Finally, values didn’t even use all 8 bits available. They threw away half the range during quantization so 8 bit computers wouldn’t have to worry about those pesky signed operations. So 7 bits is all you get, rock on.
Technologists are often drawn to music but seldom know what to do once they get there. Follow the path of recorded media: from three-dimensional wax cylinders to flat circles (78 RPM) to smaller, slower discs (45 RPM) then finally “long playing” albums (33 1/3 RPM). Or with magnetics: from 2″ wide tape traveling on precision guides at 30 inches per second to 4 tracks spread across 0.15″ tape wobbling in the dashboard of your car at 6% of the rate it was originally recorded.
This isn’t just bean counting and mass production: everyone from Ray Dolby to the LearJet guy grafted on technology and celebrated it as genuine advancement. I think they even believed it. But while learning what magnetic tape actually does on a hedonic level we found ourselves heading in the opposite direction. And even crude formats like LPs and Cassettes are making a comeback versus what we have now. Curious.
It’s not nostalgia! There’s a fundamental impedance mismatch between the digital stuff we’re handed and what people actually need and crave. Musicians, filmmakers, gamers – they’re all trying to tell us something.
Forget the great industrialists – even the modest ones were not afraid to build factories, let alone tools. Hammond made clocks, robotic card dealing tables, and musical instruments then gave music performances in the Ford Pavilion at the 1939 World’s Fair.
The Minnesota Mining and Manufacturing company started in 1902 with a goal of mining abrasives to manufacture sandpaper. Failing at mining, they bought the abrasives instead. The floor of the factory collapsed under the weight of the first raw goods delivery. Two decades later they noticed people using their sandpaper at the Ford factory were having trouble painting cars. So they invented masking tape.
Then they stuck magnetic particles on tape and designed what many regard as the best recording equipment ever, probably because many of the people working on the project were also musicians. Later, the 3M Orchestra (!) of over 100 employees recorded instrumentals for the Cantata 700 – the largest audio cassette ever produced holding 3 times the number of songs of the original iPod. It quietly generated profits for decades.
This isn’t Bell Labs level stuff. This model is totally reproducible by mere mortals in 2021. C’mon, they had a friggin’ board game division!
Laurens Hammond: “A good field for the inventor is toys for adults.” Cool, let’s make it fun.
Exactly 100 years later and here we are: terrible input devices that actually injure people. $1000+ handhelds with 450 ppi screens yet somehow we still see pixels. Basic signal processing and even gamma that isn’t handled correctly from browsers and games up to high-end A/V software. $300 round thermostats with 2.5 star reviews right next to the $19 one we all grew up with 5 stars. $500 proprietary headphones that can’t receive CD quality digital audio. A $100 digital pencil soon to be in its third tech generation. And still working on that 3D glasses thing.
What’s missing? A little bit of that…
Analog Magic (Inc.)
The problem: digital computers were invented to perform accounting and other numerical operations. If instead they were designed to support creative tasks they would be quite different in structure. (Dan Sandin, 1973.)
We are going to fix this. Input devices. Output devices. Processing tools in the middle. Automated assistants that actually provide assistance. Cloud services that actually provide service.
Why? Nobody wants to go back to work yet nobody wants to go to the movies, either.
“Platforms are collapsing. Distribution is as unbounded as it is broken. The market for consumption is larger than ever.”
You’ve heard that all before. But somebody has to make the stuff and tooling either doesn’t exist or is hopelessly stuck in the 1980s. Yes, even the Apple stuff.
Where? Primarily Atlanta, GA; intermittently Los Angeles, CA; hopefully Wellington, NZ.
Who is the target market? Everyone. If you make stuff, which everyone does, we’re going to make making it better.
No really, what is the market? Unbounded, because all tech that interfaces to the real world is broken. (We’re not surprised that Google Voice never got around to detecting hangup calls and instead emails you a loud click with a blank text transcription then lights up your phone with a notification. But 15 years later when Twilio goes public and hits $60 billion and does the same thing? C’mon already.)
Is this a media company? No. Media is interesting but we’re a maker-for-makers company and makers ourselves. We build tools. Then we build tools with those tools. Repeat.
Is this a music or video or “art” company? No. Musicians and filmmakers and artists are interesting but just like every company was forced to become a tech company to survive, now every person is forced to become a media producer just to survive on social media and video calls.
How do you constrain this?
We derive product inspiration from time-proven tools that already exist in the real world.
- You don’t need to read Dieter Rams (or nick his designs) if you remain physics-based and reality-based.
- Building tech products based on historically proven concepts and existing physical designs greatly accelerates time to market (30 products in 3.5 years in a previous company).
- Unlike the competition, we do not waste time ornamenting our tools with purple unicorns, red antelope, or pink flamingos.
We do not create products that require latency in order to function.
- Either the task is performed below the range of human perception or to an existing analog standard.
- Anything else simply destroys the product or experience, no exceptions.
- This restriction alone eliminates an enormous amount of screwing around in R&D.
We solve problems in the proper domain.
- An analog to digital conversion, a Fourier transform, three lines of code, an inverse Fourier transform, and a digital to analog conversion is one way to solve the problem.
- But a three cent resistor and a three cent capacitor can eliminate both a three dollar microcontroller and three person-years worth of software.
- Machine learning and frequency domain for analysis and open-loop feedback only; generation or resynthesis via systems based on traditional analog or physics-based models.
We build systems that work for humans.
- A useful tool performs a task and never interrupts the user from that task.
- A tool or system doesn’t need to disrupt an entire industry to be considered useful or worthwhile.
- Jidoka: “Automation with a human touch.”
Are you building a platform? A non-goal and a greatly misunderstood concept in 2021. But given our systems expertise it wouldn’t surprise us if one arose organically.
Are you building tools or are you building services? Yes.
What’s the craziest thing you want to build? Besides the card game division? Random scan displays.
Who needs that? Everyone, but games alone justifies R&D.
Do you think there’s a market? Nvidia has five different names for technologies hacking at the problem and none of them works.
Do you need to build that to be successful? No, but it would be great fun.
Can you give me an example of an actual product? Tape-based analog audio reproduction is mathematical differentiation over magnetic flux. We analyzed dozens of classic recording machines and tape formulations at the raw physics level. Then built models and developed programs to replicate all aspects of the system.
Do you need a super fast Nvidia GPU or cloud service to run something like that? No, in fact it doesn’t even require a digital computer. As such, it replicates the original system free from latency or quantization effects.
Can you give me an example of a more traditional product?
After record companies finally found the cheapest platter still acceptable to consumers, the final insult was doubling the bandwidth required in order to support Stereo. Two ears, two speakers – what could possibly go wrong? Well, one groove.
Forcing Stereo into the existing technology was so complicated that an entire sub-industry formed around the problem. None other than military giant Fairchild got into the action, building a 100 pound tool containing 11 transformers and 20 vacuum tubes. A new class of “mastering engineer” was required to operate it: controlling a postwar German cutting lathe while simultaneously twisting Fairchild knobs to limit the lateral and vertical forces that would throw the needle out of the groove.
During the golden age of music studios, taking an impeccably-recorded analog master and doing the least damage possible transferring it to a reduced-bandwidth medium seems like a feasible task for a trained engineer plus a tool. Especially when that tool is built by an aerospace company (grounded in physical concepts like gravity) with bonus expertise developing analog computers for gun sights as a wartime side gig.
But 60 years later the notion that a human can do this on music recorded by millions of amateurs in environments ranging from bedrooms to the finest studios and somehow make it sound better to everyone, everywhere listening on everything from CDs to Spotify streams to laptop speakers to earbuds to YouTube?
Well, that seems like a bit of a stretch. Some additional tooling might be useful.
How long would it take to build something like that? All done. But there’s a DJ in Orange County who thinks it could use a little more bass, so I suppose these things are never quite finished.
