Why do the most valuable NFTs look old-school and how will they look in the future?

João Abrantes
5 min readNov 19, 2021


After reading this post you’ll understand why the most valuable NFT art looks old-school and how will the looks evolve in the future. But, before we can learn that, we need to find out what made old-school art look the way it looks.

Every byte needs a good reason to exist

The computer that took men to the moon had 69.12kB of memory. To get to this capacity, half a mile of wires were manually threaded through small magnetisable rings, with each ring giving 1 bit of storage. This was crazy expensive, time-consuming and error-prone. Back then, if you were a byte and you wanted to go to the Moon, you better have had a very good reason to be in that memory! Each byte counted, each one was precious.

Rope core memory as the one used in the Apollo 11

Memory capacity was the limiting factor of computers for many years, and the best software and games were the ones that used all the smart tricks to make each bit count.

The Prince of Persia on Apple II. Pixel art was a thing, and there were plenty of smart tricks involved in the creation of these images.

Fast-forward, the price of storage drops exponentially, the art of displaying interesting images with low-data is lost and today we take hundreds of 2 MB photos of our cats being cats.

Note, the y-axis is logarithmic and the year of the Apollo mission (1969) is not on the graph. Graph from Komorowski 2014 (mkomo.com).

Why do the most valuable NFTs look old-school?

Well, this was before the blockchain had arrived. After years of disrespect and frivolous use, today bytes are precious again! At the time of writing (10/11/2021), every byte stored on the Ethereum blockchain costs around $0.3 to store. So, if you want to immortalise a 2 MB photo of your cat that will be $600 000 please.

Wait, What?! What about all the high-resolution images/animations/videos sold as NFTs? Aren’t they all stored on the blockchain?

Yes and no. NFTs use the safest public database to tell everyone who is the owner of what. They do a very good job at storing the who, but, when it comes to storing what people own exactly they simple store:


..a freaking link? That link was 31 bytes. Which is a lot cheaper to store than the actual image (it costs $9.3)! But, it is also like writing down a certificate that says “Mr. Abranti is the only and undisputed owner of the art that is on Building 31 in the Street of Manchester” now lock that certificate on the most secure safe in the universe and leave the art at the Building 31 where whoever as access to the building can change, damage or destroy it.

Some projects are storing the art on safer buildings (on IPFS for example). This is an improvement, but the certificate is still being stored in a safer and more permanent place than the actual art. Some other projects are storing a small seed on the blockchain and the algorithm that renders the art on a building. This makes little sense, the seed is useless without the algorithm that knows what the seed means. And now, to preserve the art we need to keep the two things safe…

A small group of artists, store the actual art in the blockchain along with the certificate. This is possible because these artists know how to treat bytes as the precious things that they are.

Two examples of this are the blitmap and the nouns projects.

(Left) The most expensive Blitmap sold to date (222 eth or 1.05 million dollars). (Right) Noun 63. Sold for 160 eth (730k dollars).

They look old-school because they use similar tricks to the ones used in the old days where bytes were precious. Each blitmap is a 32 by 32 pixel image with a 4 colour palette and takes 268 bytes in space ($80.4). Nouns also use a colour palette, and additionally, they use a lossless compression algorithm (RLE). Not all nouns have the same size, the noun 63 uses 455 bytes ($136.5) as an example.

These images live directly on the blockchain, and besides the safety benefits, the art is now also visible to all the code that is also stored on-chain (the smart contracts). Because of this, it was possible to combine two blitmaps to create another one and pay the creators of the original art while doing this. This was only possible because the art was accessible to smart contracts.

People could combine two original blitmaps to create another one: with the patterns of one parent and the colour palette of the other parent.

What does the future hold?

In the last years, where bytes were cheap, we’ve learned a whole lot about images, information, perception and compression. We can now use modern tools to search for the art with the highest beauty to bytes ratio.

I think that’s where NFT art is heading, and so this is what I am currently pursuing at polys.art.

The project’s goal is to use modern tools to find the most beautiful images that use low-data.

Regulars, our first collection, describe images not in terms of pixel colours but in terms of regular polygons positioned on a canvas:

Each regular takes 365 bytes at maximum. They are in scalable vector format, and therefore, can be scaled to any size without loss of quality.

Since they are entirely stored and rendered on the blockchain, they allow breeding and derivatives:

An example of a breeding mechanism, the polygons of the original are replaced by their inscribing circle. The creator of the original gets paid for the breeding.
A breeding mechanism similar to blitmaps’, one parent gives the polygons and the other gives the colours.

Since the data is on-chain, people can use the data to create derivatives in a decentralised manner.

Any mathematical transformation can be applied to the polygons to create interesting animations.

In the future, I will be posting exactly what we’ve done to create these NFTs. If using modern tools to search for beauty sounds interesting to you join us, on this search (we’re on Twitter and Discord)!

Thank you for reading.



João Abrantes

AI researcher into reinforcement learning, complex systems and collaboration tech.