THIS FICTIONAL CEO OF A DEFUNCT DOTCOM IS ALSO AN ARTIST AND WRITER
A CONVERSATION WITH JAMES LANGDON
Gilbert is CEO and Head of Derivatives at Netmorf. Netmorf was a real mobile browser company that went under in the 2001 dotcom crash. Gilbert took over the IP by participating in artist Simon Denny’s Dotcom Séance project in 2022.
Gilbert tasked an “employee” — artist and book designer James Langdon — with priming the Netmorf relaunch in the form of a novel: a corporate world-building exercise par excellence.
Drawing inspiration from early AI aesthetics, corporate graphics and medieval illuminated manuscripts, Langdon is painstakingly rendering each word as an individually AI-generated image made with Netmorf’s models on TITLES. Hundreds of these words are then arranged on each page.
The result is a fictional novel about a real company, written by a fictional CEO, and designed as an AI-generated illuminated manuscript, that explores the relationship between text and art, labor and technology, fact and fiction.
TITLES sat down with James Langdon to talk about the project.
How did the idea for this book come about?
It started with an NFT project in 2021 called Dotcom Séance by Simon Denny, done in collaboration with Guile Twardowski and Cosmographica. They were reviving businesses that went bust in the 2001 dotcom crash through a set of AI-generated logos fed on prompts describing each company. What was novel about the project was that if you bought an NFT in any of these collections it came with an Ethereum Name Service (ENS), meaning you could define your “role” as an employee of that revived company. Quite early after it launched, people were discussing the idea that if you bought the CEO token of one of the companies, it could be interesting to try and turn it into a real company.
A year after that project launched, Gilbert bought the CEO token of this company called Netmorf that had originally made an application which let websites automatically reformat their content for early mobile phones. Gilbert had the idea to publish a series of short fictions online as a way to reanimate this company.
What was the inspiration for these fictions?
There’s a wonderful book from the 1930s called Sugar in the Air about this unlikely scientific invention which synthesises sugar. The technology has really positive consequences for the world, but its development is bogged down by corporate politics and a greedy board, and eventually doesn’t come to anything.
We started adapting that novel to Netmorf, rewriting chapters of that book with new characters and loosely based on what the original Netmorf actually was. We published six or seven chapters of that, but when Mirror closed we decided to do it as a book instead.
What informed that decision?
Last summer, I moved to the countryside in the UK, close to a city called Hereford. The cathedral has an amazing library called the Chained Library, with an extraordinary collection of unique medieval illuminated manuscripts that you can request to view.
I got interested in this idea of how there are particular locations on the pages of these handmade books where labor is concentrated. It normally has to do with the handoff between scribes. You might have one scribe that does the main text and another more specialist scribe who does the illuminations and illustrations. And they obviously don’t work on the pages at the same time, so there’s a visual language of defining space for others to work into, and some negotiation over how much space each of those practitioners can use.
The discourse around AI image generation as a labor-saving technology — the assumption that AI means the removal of the opportunity to labor for creative practitioners — gave me the idea to actually use AI in a labor-intensive way, and use it to generate more. Assuming that the reason why medieval scribes concentrated their most elaborate labour on small sections of the page was because it would be too laborious to illuminate every word, I used AI to enable me to have the entire page be visually intricate.
How many pages is this book going to be?
I think it’s going to be around 200 pages.
And how many have you done?
I’ve done 10.
How long has that taken you?
It’s been about six or seven weeks of doing it — not all the time — but quite an unreasonable amount of time. It takes about 6 hours to do one page.
Tell me more about your process and how you’re using TITLES to do this.
The first model Netmorf made with TITLES was called Netmorf Corporate Graphics. It’s influenced by a nostalgia for how garbled image generation was in 2021 and 2022. The training art for that model was deliberately quite incoherent. And there’s a typographic aspect as well — fragments of words set in the Netmorf typefaces we had been developing appear in the training art. I liked that there would be some kind of ghostly glyphs in the generations.
We recently made another model, still in beta, which is called Netmorf Preferred Terms., That was trained specifically on images of text. The training data was a set of the words that we most liked that I had generated with the original Corporate Graphics model. This is the model we’re mostly using now to make the book pages.
I use text-to-image generation to prompt these models to produce words. The prompt is “a word that says …” and then the word I want in quotes. If I get a word that I really like the look of aesthetically, I’ll use that as a reference image, and I’ll shift from text-to-image to image-to-image to generate similar forms. I might generate eight versions of a word before I get one that I like.
Sometimes the words are on the edge of being readable, but you can generally deduce the meaning in the context of the sentence even if it’s visually unclear. Especially with longer words, you’ll often get something a bit garbled. For example, this sentence should read, “The agent had underestimated the fragility of Render’s headspace.” But the generated word actually says “underestitiated.” I’ll just accept that because I think it’s readable in context.
Sometimes I’ll generate blobby forms to define word-sized shapes on the page, and then work with another tool that a friend and Netmorf colleague built for me, Simon Knebl, which is for setting type into irregularly shaped forms.
It’s a very Joycean project — both in labor and in how different sections of the text go in and out of readability. Do you intend for this book to be read, or engaged with in another way?
I do really care about this text and I’m invested in using design to imbue it with some Netmorfish essence which I hope will persuade people to read it
I think a relatively mainstream anxiety about generative content is that consuming things made by other people will just become irrelevant because we will all be so enthralled by our own personalized creations. A few years ago, I saw this post on X that said something about how nobody will read text anymore, in the LLM age, but that “designer paragraphs” could still cut through and get attention on merit. I am just really fascinated with that idea of the designer paragraph.I’m taking it a bit literally, obviously. The OP didn’t necessarily mean to apply more visual design to a text, but it’s this idea that some text will be imbued with something extra that will make it a meaningful signal for other humans.