About Carecore

Carecore is a web page where visitors can plant, feed, treat, and compost seeds. It was crafted by Raphaël Bastide, for The HTML Review .

It has been made without AI assitance but with the help of Drice Ducongé dos Santos.

Visit the Rules to learn how to interact with the seeds.

This project has a singular story I tell in Process below.

Process

The project Carecore was not supposed to be the floral composition you can see on this page. Initially, I submitted a totally different idea to HTML Review. The idea was to develop a web app allowing mixing drawing and singing, that is what were drafted and sent to the editorial team. Then, a couple of weeks after the submission, this happened:

I casually opened an app I didn’t know in a new tab. As I understood, it was another generative AI model that outputs images. Even if I don’t use those tools for my personal practice, I take a look at some of them from time to time, so I can follow what happens in the digital world and remain up to date with recent technologies, for my students, and for myself. I have recently realized that those tools have an unpredictable effect on my practice, and you will understand why in this text.

Then the tab was opened in an isolated container so I could log in without leaving a trace. A common prompt interfaces shows and I type:

“Circus Poster with big title: ‘Olivia’s circus’ featuring circus animals and clown faces, not too nostalgic style, nice typography, modern and dynamic composition.”

Then I waited and saw in a sidebar, that files were being created: “index.tsx, package.json, App.tsx…”. I was surprised, the prompt didn’t trigger the creation of an image but a web app. My surprise continued as the pile of files and folder kept on growing. I could identify the use of Typescript, Node and lots of dependencies. I personally tend to avoid that kind of structure for my personal projects, but I was too curious to see what will result of all those code being created under my eyes, from a prompt that was intended to be for an image. After 20 second, the app appeared in the main area under my request. A colorful poster with a bold title “Olivia’s Circus” appeared, illustrated with pretty common circus imagery, but that was not all. The poster was framed with a whole interface with its own logo titled “Poster Magician”. This interface actually allowed users to customize the poster: the main title prefilled with “Olivia’s circus”, the font choice, and there was even a prompt text area. I couldn’t believe what I saw: I prompted for an image in an AI app maker that decided to include AI by default in the App I didn’t ask for.

Surprised but curious, I was still impressed by the technical ability of this LLM chat interface. I decided to prompt an actual app idea this time, something more strange, something I had in my head since a couple of weeks: the idea I submitted for the HTML Review, and that has been accepted by the editorial team. So I write down:

“A minimal drawing tool where every stroke is a sound loop. Drawing is recording: mouse down starts recording, mouse up stops recording and starts looping what has been recorded. Move the strokes to mix: X axis is the pan, Y axis is the volume of each strokes. Don’t use node nor dependencies.”

30 seconds later, a big amount of files, not minimal at all, had been created. An app appeared, very much based on several node dependencies and unwanted typescript files. The interface was very predictable: dark gridded background with serious and technical fonts and composition, looking like some famous ChromeExperiments from the 2020s. But I decided to test it anyway to see if my idea is functional and / or fun. I moused down, vocalized, moused up. Wow, my voice was looping straight away. I pushed again my index finger, tried singing 2 tones above to create a simple polyphony, and released my click. The two tracks played together with different length, creating a basic but interesting polyrhythm, that is what I expected. I moved the tracks around, playing with left / right panning and voices’ volumes. Quickly I found this live performance quite fun, but even more quickly, I got a bit bored with it, strangely.

I felt alone in front of my screen, playing with an AI generated program I came to understand almost nothing about. I knew nothing about its guts, its bones, its diseases, its imperfections, its weight, its limits, and the way it was born. At that time, the experience I got was limited by the odd lack of soul of that program. Everything under my eyes looked and felt like déjà-vus: the interface looked like a predictable product, with over-generic graphics and typography, that could have been used for any kind of interface. There was absolutely no space for any expressive choice, this shell was empty. I could use my voice the best way possible, create the nicest loops, the result would eventually be disappointing because this app was just another app. This LLM transformed a singular idea into a generic and boring result.

In the following weeks, I thought a lot about this experiment. For days, I had in my head the following questions: Should I keep and modify this app, so I honor my initial submission text to the HTML Review? Should I re-create something functionally similar, but from scratch? How is it possible that a technology I reject has such an influence on my creative process? But the fundamental question was: What is my position now?

Caring means being able to synchronize. We can’t care about something we don’t understand, we can’t take care of a black box. Our ability to help, support, empathize and raise, assumes that an alignment is possible between us and the subject of care, implying that subject, person or project, is not fully stranger nor opaque.

I recently realized that a group of artistic projects I made the last ten years were about taking care, or giving care to things.

• Twins (2016) was a 72 hours performance originally intended to develop two artificial intelligences. Through time, physical and psychological exhaustion, I ended creating various artworks about digital paternity and parenthood, and what does it cost to give “life” to something, idea, artwork, artificial being. I remember wanting to feel, physically, what was the effect of putting myself at the service of something I created, and ultimately that the result of my project was more for an AI public than a human one. At that time, I was not considering becoming a parent (of a little human), but that performance left a mark on my body’s memory that I recalled 10 years later during challenging first months of parenting.
• Nest (2024) is a web based artwork made of salvaged visual and sound elements, combined with non-linear musical composition, adaptable to visitor’s interactions on HTML audio players. Nest’s focus is the preparation of a sensible and cozy environment around a MP3 file I inherited from another artist. I remember realizing pretty late in the project (once again, awareness came with time and distance), that its relation to my soon-to-be-born daughter was not a coincidence.
• Being a script (2025) is a short text I wrote during the first months of my daughter. I had no conscious intention of writing about being a parent, but it has become evident that this topic reveals itself between the lines: In Being a script, Deval, a short program introduces itself and its relations to the world, especially the hands that created it, but also its own history and how its time is different from ours. The text shows how fragile and sensible the program is, but also emphasis on Deval’s (a non-human cultural object) wish to live and feel.

That line of projects led me to developing Carecore, still in the making at the time of writing this text.

Carecore was crafted in reaction to the drawing-to-voice app that the AI model called “Sonic Sketch”. The first versions of Carecore were some CSS floral ornaments, able to evolve with time and small interactions (feeding). It was a relief to make something fully by myself, with my own tool set and sensibility, understanding every function, every graphic choices, working on small details here while leaving big bugs there. That feeling led me to more radical choices: I wanted to integrate in that project, elements that felt extremely singular, wonky and very expressive. I wanted to make as much choice I could, and push every one of them to its most viscous state. I often chose to go in the “un-promptable” direction, I felt good each time a choice had no written language equivalent.

Then, the project Carcecore evolved as a collaborative laboratory, where flowers pieces could be created, fed, treated and composted. I carefully avoid it to become a game as there is no goal and nothing to win, no dopamine to chase. But the project, I hope speeks by itself, about where it comes from and how it evolved. I want it to reveal both its imperfections and its forces, how fragile and flamboyant it can be at the same time.

I have no distance to judge if Carecore is a “good” project yet. I can’t see yet how it will be used, how it might age, and where it might continue to evolve. But I am definitively already attached to it, because it strengthens my appetite for crafting things myself, making choices, a lot of them and above all, not always the good ones. I am happy this project revealed a line of creative motivations about a certain form of care, that have never seen clearly before. I am grateful I chose to take a turn from my initial idea because I learned way more about myself on that new path. For that reason, that project is maybe already a success.