Feature montage by Takens Theorem / with permission from j1mmy.eth

NFTs Are Adaptive Entities

NFT projects reflect their network niche, and they are evolving

Takens Theorem
Coinmonks
Published in
10 min readDec 2, 2021

--

As a concept, “meme” was born in the late 1970’s, in the famous book The Selfish Gene. Described by analogy with genes, the concept of a meme is a unit of cultural and behavioral selection. Memes are transmitted from person to person — a hairstyle, a song, a way of saying something. When a meme takes hold of you, in some sense it has “parasitized” your nervous system; your brain and behavior are part of its niche:

From The Selfish Gene (1976)

Andrew Steinwold has referred to NFTs as new forms of “monetizable memes.” In a visionary Zima Red post about this idea, he uses this meme concept as literally and sharply as its original formulation: “The most exciting part about monetizable memes is that this is just the tip of the iceberg. We are going to see these assets evolve in all types of ways we can’t quite understand yet.”

Andrew Steinwold’s tweet.

There is something compelling about this idea intuitively — NFTs are, after all, part of our current digital culture. So it seems reasonable that they, too, are subject to similar memetic analysis. But Andrew’s implication above is that NFTs are emerging from a new substrate — the crypto landscape creates a decentralized, uncensorable public ledger. And with EVM chains in general, it will be possible for NFTs to interact and build upon each other in an open, composable computational ecosystem. The NFT space produces “living structures” that may adapt, evolve as they further parasitize the chain — and us.

The “DNA” of an NFT

This brief post illustrates that we can already see hints of this evolutionary dynamic in the chain itself. One recent example fits in this analysis nicely. Loot is a simple SVG image of a list of game items, echoing classic Dungeons & Dragons and other fantasy gaming (see token #1 here). Its simplicity made it easy to copy. The Loot NFT contract was copypasta’ed repeatedly into new, propagating “mutations” that choked the network. The More Loot derivative was practically uncapped, and for a stretch of time this past September (2021), it had a transaction in almost every block.

Loot and its derivatives deploying onto Ethereum.

This example suggests that NFT projects are indeed memes in precisely this evolutionary sense. Loot literally mutated and continued to propagate. The “niche” here was the minds of NFT enthusiasts who shared this excitement — a project that was conceptually and technically interesting, with warmly familiar vintage themes.

Loot illustrates another idea that classifies different NFT projects. Loot is an on-chain NFT — each Loot SVG image lives inside the contract. If creator Dom Hoffman vanished along with Loot’s website and all that remained was Ethereum’s ledger, chugging along, it would still be possible to recover your SVG in the contract’s tokenURI() function, along with all the traits of your “loot.” There are now many such “on-chain” NFTs. See a list here and here.

So this evolutionary framework goes deeper than simply studying the conceptual, cultural properties of Loot. These properties could be considered its “phenotype,” its directly observable characteristics. With on-chain NFTs, and their sometimes elaborate contracts, it is possible to study how programming strategies may also have these evolutionary properties. This also has a playful biological analogy. Consider Andrew Steinwold’s note about surprising and new evolutionary properties. By studying these contracts, we can investigate adaptive characteristics in the “DNA” of these memes — in their underlying contracts (their “genotype”). Consider two examples.

From Memory: Avastars

Avastars is among the first fully on-chain NFT art projects. The artist Marmota vs Milky created an elaborate, high-definition set of SVG features that can be used to compose a massive space of possible avatar profiles. In collaboration with nft42, they released this as a beautiful NFT project in early 2020. Across 2020–2021, many fans of these Avastars became addicted to scrolling their website, to mint the permutation of features perfectly suited to the avatar they desired. And just this month, Avastars released the “Replicant Factory” — prior owners can now recombine features of their NFTs to create new, unique Avastars: Replicants.

When looking into an Avastar’s “DNA,” the contract, we quickly see that this NFT project is partly a function of a particular niche — the network conditions of Ethereum at the time. Gas prices could be as low as 5–10 gwei, and ether was hovering around $200–300. Etherscan can reveal how the project was released, and we see that these beautiful SVG layers were encoded directly into memory on the contract:

Deploying SVG features directly into memory. (Etherscan)

At the time, the main Avastars contract was deployed for a gas cost of $10. Its entire suite of features, over 500 of them, for less than $2,000. But Ether has increased in price since then, and base fee of Ethereum at the time of this writing is 10 or 20 times higher. The cost to release the contract alone would probably be about $2,000. To release the whole batch of artistic features, the project might cost over $300,000.

When building an Avastar, the contract takes these layers and assembles them into your piece. It’s 100% on chain — the renderAvastar() function can be used to recover your SVG image.

Comments can be fun: “assemble SVG sandwich.” (Etherscan)
Avastar #14912

From Computation: Squiggly

Other memory-heavy approaches can be seen in innovative early projects like CryptoSketches, which stored an entire art animation in an ERC-20 deployed in early 2018. Some have also used the calldata in a transaction, like the mysterious neural-net generated monsters 0xmons. This project stores the encoded NFT inside a transaction’s relatively cheaper calldata.

CryptoSketches from 2018 by Joey Chips. 100% data on chain, even the dynamics.

But there is another strategy. The strategy is to use computation more than memory — to encode an NFT by how it should be constructed, rather than having its parts preconstructed and stored explicitly.

A later 2020 example can be seen in Squiggly by NateAlex, which encodes an elaborate and beautiful generative artwork 100% on chain. Squiggly’s contract deployment faced a gas price of over 60 gwei. The contract reconstructs the artwork each time by seeding it with a unique number set at mint. This reconstruction is done by a contract call — it is free. In fact, because the pieces are generated by computing over a seed, Squiggly has a public function renderFromSeed() that can be used to explore a “synthetic squiggly,” what your piece might have looked like. This approach defines a function which computes the piece each time.

Drawing Squigglies, inside renderFromSeed() (Etherscan)

I recently experienced this transition in strategy. Two projects, the_coin and Gaussian Timepieces, are distinguished in this way. the_coin pieces encode Bitcoin history as a data-driven visualization, stored in memory, and owners can modify them with CSS — it was deployed at 6 gwei gas price. Gaussian Timepieces are dynamic SVG animations representing time on chain. To overcome the spike in gas issues this fall, these timepieces are constructed by a set of assembling functions, modified by using token ID and owner address as input (in the spirit of Squiggly).

Squiggly #66

The Hypothesis

The above discussion poses a hypothesis. It is not just that creators adapt to the network. This is obviously true (and still interesting to study). Instead, the hypothesis is a little deeper, that the adaptation of NFTs and their contracts can be understood with an evolutionary lens — projects as a whole are adaptive entities, constrained by the external world of creators and collectors and market forces, but also by the chain itself. We see this in the “DNA” of the project, how the contract is assembled to achieve particular ends.

There are many further examples that could be considered here. These other projects would show that this distinction between memory and computation is not so clear cut — projects can balance these demands to come up with creative solutions. This makes sense again from an evolutionary standpoint — the traits of an NFT project are complex, and there is much more fluid adaptivity. Examples include the bitmaps of Brotchain, the 3D renderings of Strange Attractors, the dynamic cursors of Blitmapeven pieces that leap into the world like Moon in Motion, which updates a visualization of the moon’s phase based on the on-chain timestamp.

Should we consider these features of an NFT project when appreciating it? For some emerging curatorial frameworks, these technical details may be a key feature of a project’s allure. Consider Avastars again. This technical characteristic of Avastars makes it fascinating, and being “early” is not just about something coming before something else — early can also mean qualitatively different in an underlying and interesting way. Avastars are “heavy” NFTs, their contract loaded up with elaborate artwork on a network that was once more permissive in fees and ether price. It has that austere weightiness of artifacts from the past, lasting forever into the future, to inspire awe in those who float down the chain and inspect the contract’s details.

Floating past Avastars on the chain.

Under the present resource limitations, some NFT projects have simply wandered into new habitats. The #CleanNFT movement has led to significant creation on Tezos. EVM/PoS chains like Polygon (Ethereum side chain) and Solana and Avalanche have attracted some development. Projects on these platforms are under quite different resource constraints, and it may be interesting to compare the NFT “DNA” across these ecosystems.

The future may get yet more interesting. We may have AI infused with NFTs, and adaptive processes can be yet more complex. The projects of Alethea AI are exciting, and they fuse on-chain provenance of NFTs with external servers required to train and deploy AI. Think: ownable chatbots, companions. Such complexity cannot be encoded on chain for a variety of obvious reasons. However it may be possible for small-scale snippets of AI to be encoded on chain, to be used and assembled off chain to create new variants. An amazing recent project ArcadeGlyphs by Inner Space and Captain Pixel implements a simple AI to play an arcade game on-chain, producing a dynamic of an NFT’s performance as the final visual representation.

Conclusion

There are exceptions to the above hypothesis. Autoglyphs, for example, deployed at 6 gwei fees in 2019, but used the reconstruction approach in its draw() function. But this violation of the hypothesis isn’t surprising. Features that are “selected for” in evolutionary processes are often already attested in a population. Their frequencies get amplified or dampened in the face of changes in the ecosystem, in the niche. The Avastars approach is an incredibly rare and wonderful illustration of the alternative, back when the network permit it.

The evolutionary metaphor departs from these scientific theories in some ways.

For example, these NFT projects are literally the product of an “intelligent designer,” a notion that is avoided in evolutionary explanations. The above patterns in Avastars and others were literally designed. But their properties can still be seen as a resonance between organism (NFT) and environment (network, etc.) made famous long ago by predictions of Darwin.

Darwin’s predictable moth (Wikipedia)

There’s another distinction between NFTs and biological evolution, in particular fully on-chain NFTs. It’s an eerier one. In the original evolutionary formulation, behavioral memes can of course die, like biological organisms. That’s a key part of the whole evolutionary dynamic. An NFT project can lose attention, and sit quietly on the chain without engagement. Is that “death”? NFTs live permanently but often functionally on public blockchain. They are “living fossils.” When they are dug up, they can spread again, in a new niche that may be more auspicious. They are, in this sense, immortal.

--

--

Takens Theorem
Coinmonks

Dynamic distributed data displays. Intermittent. Friendly.