If you are a regular Artnome reader, you know we are big on blockchain and generative art. So of course I was super excited when my good friends Matt Hall and John Watkinson of CryptoPunks fame gave me a sneak peek of Autoglyphs, their new project which creates old-school generative art that literally lives on the blockchain.
In this post I nerd out with Matt and John about Autoglyphs, grilling them with all kinds of questions including:
What are Autoglyphs and how do they work?
How do Matt and John manage to actually put art on the blockchain?
Did early generative art serve as inspiration for Autoglyphs?
Why did they create just 512 out of billions of possible Autoglyphs?
What are the differences between Autoglyphs and CryptoPunks?
Do Matt and John think of themselves as artists?
What makes a good Autoglyph?
Autoglyphs are highly unique because traditionally, the actual image files associated with blockchain art like CryptoPunks, CryptoKitties, or Rare Pepe are stored in a database somewhere “off chain,” meaning off of the blockchain. Artists typically address this “off chain” storage by including a link to the image from the blockchain called a “hash” so that you can locate the image file for your artwork from its record. For example, even though the image of my CryptoPunk is comprised of relatively few pixels, it actually lives “off chain” on the LarvaLabs server at:
This means the actual artwork does not technically benefit from any of the tamper-proof advantages like “decentralization” or “immutability” typically associated with the blockchain (unless you think of the token itself as the artwork instead of the image). Put another way, there is nothing stopping someone from altering, moving, or removing the image from the location the hash is pointing to. If that were to happen, all you would be left with is an immutable record stating that you own an artwork, with no way of actually seeing it.
Perhaps you are thinking, “Why not just store the image on the blockchain? It is, after all, a database, right?” Well, blockchain is great for a lot of things, but storing large image files is not one of them. Unless you can make art with a super tiny footprint, it is impractical to store traditional image files like JPEG or PNG on the blockchain.
This is what makes Autoglyphs so damn cool. Matt and John decided to accept the storage limitations of the blockchain as a challenge to see what they could create that could actually be stored “on chain.”
I love this idea because it is a throwback to the compute and storage challenges that the earliest generative artists like Michael Noll and Ken Knowlton faced when trying to create art on computers in the early 1960s. As you will see, this is not lost on Matt and John, who are huge fans of early generative art and decided to embrace the aesthetic and run with it. With that, let’s jump into the interview.
Autoglyphs - An Interview with Matt Hall and John Watkinson
Jason Bailey: Thanks for chatting, guys. I have a bunch of questions, but I’m happy to start with the easy one. What was the impetus or inspiration behind Autoglyphs?
John Watkinson: There is a lot of talk of art on the blockchain. With the CryptoPunks, all of the ownership and the provenance is permanently and publicly available, and those rules are set and fixed. And yet there's still a bit of an imperfection there in that the art comes from outside of the blockchain and stays out there, and it's just referenced by a smart contract. We don't have any complaints about the CryptoPunks, but it felt like there was an opportunity to go further. With Autoglyphs, we asked ourselves, “Can we make the entire thing completely self-contained and completely open and operating on the blockchain?”
JB: So the decision to literally store the artwork on the blockchain comes with some pretty hardcore restrictions, right? What sort of parameters are you now boxing yourself into once you make that decision?
JW: You have to have very small and efficient code generating the work. The actual output of the work has to be a very small amount of data or text because you can't have a large amount of data on the blockchain. So a small amount of efficiently running code, and fairly small, efficient output.
Those were the constraints, and they were pretty extreme. For a while thought we couldn't do it, or couldn't do it in a way that was satisfying for us. I was sort of exploring various generators and trying to make them more efficient, just binary image generators. I got to one that I thought was pretty good and I then experimented with it, trying to turn it into a smart contract, and I just couldn't get it to work. It was just hitting limits and wasn't working at all.
Then I tried it a few months later and just pushed it a little further and just got there. Still, the transaction fee of making an Autoglyph is going to be about half of an Ethereum block. So an Ethereum block is about eight million gas, so that's how much computation can happen in one mined block of Ethereum, and this is going to be three million gas, so it's almost half a block.
That means that the transaction fees will be relatively expensive - between one and two dollars - depending on the price of gas. So it's a pretty hefty transaction. If we went much more than that, we would already be outside of feasibility. If we went over eight million, it would be completely impossible, you wouldn't be able to do it.
JB: Got it. Dumb question: Does the code for generating the image live on the blockchain? Or is there actually an image on the blockchain?
JW: The code lives on the blockchain, and in fact, when you ask the blockchain for the image, it will just generate it again for you. That part happens on a end node, so that doesn't cost any actual money or gas. But whenever you say, “Give me the image for Autoglyph five", it will just generate it again for you based on the seed information that was created in the transaction.
Matt Hall: It's probably also worth making the distinction between the image and the instructions to generate different representations of it. The actual image you see on the website is not generated on the blockchain. The art, the instructions for how to write it are on the blockchain, but we make an SVG or PNG file on the web server. If that was your question, then no, the actual image data doesn't come off the blockchain, but there's an ASCII representation of exactly what that is on there. It's an ASCII art representation of the glyph.
JB: Nice. That was going to be my next question. I love ASCII art, and I assumed that it was generating some sort of ASCII format. So the ASCII art version of the image is an image made out of text and is actually on the blockchain. But in addition to that, you're generating PNGs or JPEGs for end user convenience that you've got hosted at Larva Labs? Is that a fair way to put it?
JW: Yes. We're generating the image, and we basically created instructions on how to do that. So in the source code for the actual smart contract, if you scroll down a little bit below that big ASCII art “Autoglyphs,” you'll see that there are these little instructions. For every ASCII art character, it tells you how to draw it. We generate image files that way. But the idea is that anyone can generate it - kind of like a Sol LeWitt instruction set for creating a drawing. If you own a glyph, then you can make it at any scale, with any materials you want. You can make your Autoglyph using these instructions.
JB: Great. That was going to be my next question. Is it a bit like a Sol LeWitt, where essentially if Larva Labs, God forbid, goes out of business and you decide that you no longer want to support the interface people will have everything they need built within this little blockchain code to infinitely generate these Autoglyphs? Will Autoglyphs outlive us all?
JW: Yes, that's the idea. They'll be able to make their Autoglyphs and follow these instructions to render them. We have a little pen plotter, we're going to make some of our Autoglyphs physically rendered with that, which is kind of just for fun. It's well set up for plotting that way.
MH: We were kicking around different versions of this and then we saw this show at the Whitney. It is a retrospective of a bunch of digital art. They have early generative art and all sorts of different stuff. There was this big Sol LeWitt piece, and they were explicit about how this piece had been executed by an assistant at the gallery, but that's in keeping with the intention of the artwork and the instructions of the artist. We thought that was good, it was perfect, because we can't do a lot of things we want to do directly on the blockchain, but we can have the spirit of it be completely self-contained.
By providing them with the instructions on the blockchain, now the art can be rendered very large and detailed. For example, we could have stored these as tiny pixel graphics, graphs, something like that, but then you're limited to that. This way they can operate at any scale and in any material.
JB: It does feel like a throwback to some of the early generative art. I'm thinking like Ken Knowlton and Michael Noll. Other than Sol LeWitt, were there other artists who inspired the Autoglyphs? Or do they just look like old-school generative art due to the storage limitations of the blockchain?
JW: A little bit of both. We definitely needed to clamp down the parameters pretty hard because of the technical requirements, but we'd been getting into the early pioneering digital art of the '60s and early '70s stuff. It's definitely an homage to Michael Noll and Ken Knowlton and that kind of stuff, which we really love. Only once we got to this digital art world via the CryptoPunks did we really realize how much of all this stuff had been explored in the '60s. It’s almost humbling how much ground was covered so quickly in digital art in the '60s and early '70s.
JB: Yeah, I love early generative art. It looks like from the Autoglyphs site that the algorithm, while it had to be simple by definition, is capable of generating billions of unique artworks, but then there are 512 that ultimately will be produced before it stops, right? So how do those 512 get selected among the billions of possible works? And second part of the question, why 512?
MH: Good question. They're going to be randomly seeded. There's a random seed that goes into the algorithm to generate them, and if you operate the contract manually, you can specify the seed manually - but you can't reuse an existing seed that's already been used to make a glyph. We debated whether to limit it or not, whether to make it so that everyone and anyone can come and get their glyph. There are a few arguments in each direction, but ultimately when you make generative art like this, the generator kind of is the artwork a little bit, and there's so much it can express.
It's basically a very tiny generator. If you scroll down in that source code, the core of the generator is the draw function, which is only about forty lines. So we said, “At what point does a generator kind of play itself out, where you've seen everything?” You could make more, but it's just going to be like, “Oh, that's similar to that one, that's similar to that one,” so how much surprise and variety can it really deliver? So we found that threshold.
We made it a power of two just to keep it nerdy. But that was the around the threshold where we said, “This is about the right amount of these things in order to fully explore the generator but not make them all worthless because there's a myriad of other ones similar. This should be enough to discover cool surprises and get a sense of what it can generate and have a good collection out there, but not hit it too hard and destroy all the mystery of it.”
JB: Sweet. And then you mentioned on the site that 128 of the Autoglyphs are already claimed, so who claimed them?
JW: We're going to claim those. We want to have a decent chunk that we can explore and mess around with, and we want to display them in large groups together. That's how many we're taking for ourselves and the rest are going to be up for grabs.
MH: It's a similar model to the CryptoPunks, where we wanted to convert ourselves into the same kind of relationship to the artwork as everyone else. So we just become owners after the thing is launched, and we like how that sort of played out on CryptoPunks. People ask, “Why don't you take a cut of all the sales?” Well, we didn’t take a cut of the CryptoPunks, so we want to just be the same as everybody else. We felt that that still was the right way to go with this.
JB: Right. It's experimental and you're along for the ride with the same level of risk as everybody else, right?
JW: Yes, exactly. That informs the sale price for the rest of them and where that money goes, and then we don't feel like we need to claim the sale price of those things. We can donate that because we have a portion of the artwork.
JB: Got it. No, that totally makes sense, and I'll come back to the charity stuff, too. For me, at least, CryptoPunks was sort of stealth generative art, meaning that most people don't know what generative art is, and they didn't need to in order to love Cryptopunks, right? I think part of the appeal of CryptoPunks was that anybody could look at them and get it and fall in love with them, like, “Oh, cool, look at all these different cool characters.”
You also received interest from art nerds like me and you were in that awesome show with theKate Vass Galerie. Are you worried at all that the Autoglyphs may not have the same broad appeal? Or maybe you didn't even assume that there's going to be a broad appeal for CryptoPunks, either, kind of going back to your assumption of these things sort of being experiments?
JW: Yes, I think that's what it is. We didn't expect it with the CryptoPunks; we don't really expect that here. We know people like you and the other people we've met who are into this stuff, and we know that there will be at least a narrow appreciation of this for the same reasons why we dig it. But no, we don't necessarily expect it to have as broad appeal as the CryptoPunks, just because they were a little more consumer-friendly, just easier to engage with, easier to understand. You didn't necessarily need to know that they were generative, you just liked them, like, “I want one that looks like me.” You're not going to find an Autoglyph that looks like you, so…
MH: If you do, that'd be cool!
JB: I like that challenge — that's the first thing I'm going to do when I get off the call.
JW: Yeah, it's more of like a Rorschach image.
MH: You see your true self in the Autoglyphs.
JW: Exactly. Yeah, you see your emotional self. We took the attitude, “Let's not worry about that; let's just kind do experiments that we like and we think are cool and resonate with us.” But there's no doubt that we were like, “Let's keep the size small here, because the audience might just be smaller, and that's fine.” It doesn't need to be as big or as wide a variety of people owning it or as high a transaction volume as the CryptoPunks.
JB: Got it.
MH: I think it's fair to say that we're starting to think a little longer term about these things, too, now that we're coming up on two years of the CryptoPunks launch. We thought CryptoPunks might be just a blog post, a couple weeks of interest and the end of it — and it's still going strong. And then seeing this generative art from the '60s and having some similarities with the very limited computing ability we have to work with, it just felt like, “There's cool stuff to explore here that could have appeal long term.” It's okay even if doesn't have the broad appeal at the moment, it's fine.
JB: What are you guys? Do you think of yourselves as artists, and had you in the past, or has that changed in the last few years?
JW: It's funny you ask the “what are you guys” question, because we've been looking at each other the last couple weeks asking the same question. What are we, what are we doing here? We're quite a wide variety of things, and this is one of them.
And obviously it’s almost a loaded term: We're artists now, I guess. And especially looking at generative art from the ‘60s with fresh eyes. There were a lot of people working at Bell Labs and just experimenting and trying things out. Then in hindsight we can look back at that and be like, “Man, that's really cool art that really predates this whole digital art thing.” And they were just engineers, they were nerds just expressing themselves. I think we put ourselves in that camp happily, so not claiming that we're career artists or that's what we’re trying to promote ourselves into, but claiming the ability to express ourselves and make things just like anyone else.
I don't know, Matt. Is that how you feel about it?
MH: Yeah. I felt more comfortable with that term when I found out the history of technicians becoming recognized as artists because they have the skills necessary to operate something new.
JW: And they were thinking about it more than anyone else.
MH: Yeah, just familiar with it, and would see the limitations and the strengths in how they're utilized. So I feel pretty comfortable in that category.
JB: So the CryptoPunks were initially free. Autoglyphs are coming in at like $27.69 with proceeds going to the charity350.org. Could you maybe share a little bit of the thinking behind that? Why 350.org?
MH: Even with the CryptoPunks, where we gave away 9,000 of them, a large number of them went to a few early people that just got on it and automated that process, so we wanted to avoid that. We wanted to a have a better distribution of people, so we felt like the best way to have that was some price associated with generating them .
JW: So then the solution there was, “Let's donate that money to charity,” and then if the whole set sells out, then it will be a pretty good total.
MH: So if we can sell out of these things it'll be about $10k to 350.org, which is a good organization for trying to move power generation over to renewables. It felt like the right fit in all of those dimensions.
JB: Great, yeah. A softer ball question, so from each of you, what makes a good Autoglyph?
JW: I think with a generator you kind of get a sense of what it makes and then you get surprised by a few things. So I always like the ones that are just like, “Woah, that's not what I expected.” Once you look through 40 or 50 of them, you can always tell which ones are crazy or weird looking, and it’s always fun when it kind of breaks out of expectation. Those are the ones for me that I like. I like ones with diagonal lines. For some reasons, those are the most appealing, ones that are just made out of diagonal lines.
MH: I think we both like the ones where, because the symbol sets are simple, it’s cool when you get the sense that there is a pattern there that's not actually there. There are ones that look like there are curves in them, but there aren't. I like that a lot. I also like ones that look different at different scales. So when they're zoomed out, they look like one thing, and then as you zoom it in, it dissolves. It’s something we're trying to figure now when we're working on physical representations of them, how thick should the lines be, what's the ideal viewing distance, where do these patterns resolve? I think that's my answer.
JB: Cool. And then anything you want to share on the launch process? I think you mention the date in the email, but are there plans to show the physical works anywhere specific?
JW: Yeah. We're just going to launch them first just on the web and on blockchain, and then we'll figure that out next. I think we do want to show a bunch of the glyphs that we claimed for ourselves, maybe one of the art shows in New York in May. We're going to figure out which one's the best one to do that for. We haven't totally figured that out yet. We first just want to put it up, we still want it to be an experiment that pops up on the internet and not have it be a gallery-type launch or anything like that.
JB: Thanks for your time guys! I think Autoglyphs are awesome and can’t wait to add some to the Artnome collection!