I've long been interested in science fiction, especially that which paints a positive vision of the future. I'm also passionate about artistic expression, even though I'm a software engineer. Somewhere between these two passions, in 2021, I came up with the idea to create an online club dedicated to sci-fi art in addition to my main job. It all started with selecting and posting materials, primarily from DeviantArt and, to a lesser extent, ArtStation. But with the rapid development of AI, especially in image generation, the club became more unique, as I was able to translate my ideas into art. Chat and code assistants are also used behind the scenes. AI is clearly the technology of the future, and using it in a project that promotes a positive vision of that very future seems more than suitable. I'd like to share this experience. There won't be any deep technical details about AI, this is more of an overview and presentation text.
The club's main platform is the Russian social network VKontakte, where it has over 4,000 subscribers. There's also the Fediverse page and recently launched Telegram channel. Posts are made twice a day — one by the artists and one by AI, generated based on my prompts.
The current AI use cases for the club are as follows:
Imagine Art (previously Midjourney) is used for AI art generation.
Art ideas are partially provided by Google's Gemini.
The code for the club's website is generated almost entirely in Cursor using Claude 4.5 Opus High from Anthropic.
Now let's look at each of the scenarios in more detail.
At first the article was posted here in Russian and recently I decided to translate it.
Art generation
My first attempt at AI-generated art was the "People of the Future" series. By this point, I'd already reviewed numerous artists' works and noticed one significant drawback: while there are many examples of spaceships, stations, robots, and so on, there's a noticeable shortage of images of people, especially peaceful and positive ones. The full series can be viewed in the VKontakte album; here's one example:
Subscribers appreciated the series, and I began using the generator for a wider range of tasks.
The main theme of science fiction is space exploration, and much of the art is devoted to this. Here are a couple of examples:
Robots are also present, but not as a threat to humanity, but, for example, peacefully playing basketball:
Futuristic cities, bright, clean, peaceful, and lush with greenery, are also an integral part of the project:
Nano-themes are also present:
And, of course, we can't do without time travels:
It's also worth adding that the future consists not only of heroic deeds and fantastical journeys, but also of simple everyday life, for example, in a school like this:
I wouldn't say creating prompts for art generation is super complicated. I simply use an approach consisting of the following parts:
A general description of the idea — for example, we want to depict a female engineer.
Necessary epithets, such as "adult" and "experienced," and for variety, we'll include an Asian race.
A strong focus on a positive image of the future.
Details vary depending on the mood.
So, for the prompt "adult experienced kind sci-fi Asian female engineer," we get:
In fact, providing a wealth of detail doesn't always work; the AI sometimes simply ignores certain details, so I try to keep the prompts simple and hope the assistant's imagination works well. Above are examples of both short and detailed prompts; you can check how well the setup was executed. For example, the AI ignored the dinosaur's herbivory in the time travel image.
And the clarification of the positive image is connected to an interesting feature: AI often paints the future in gloomy tones. Here's an example of "Human in exoskeleton, Sci-Fi style" with and without the "positive" clarification:
Generally speaking, the art is mostly of fairly high quality, and oddities like six fingers (though this does happen in nature) are practically non-existent.
Idea generation
I admit, I don't always have the imagination to publish daily, so sometimes I turn to Gemini for help. The answers are often quite useful, like these:
Here is an example of a picture based on one of the ideas:
But sometimes the AI just plain hallucinates and comes up with something simply unimaginably weird, like "A majestic "cathedral" ship with soaring gothic arches of white stone and stained-glass viewing ports" 😀 I certainly don't publish such nonsense.
Code generation
I have little understanding of front-end development (my specialization is Linux, Python, and QA), so the club's website was initially made very simple by me (nothing but HTML and CSS) and served solely as a primitive “business card”.
But with the development of AI assistants, I couldn't resist trying using Cursor with Claude to make the site more beautiful. As a result, without me writing a single additional line (I only prepared the content), the AI completed the following tasks:
Switching from static to React while preserving the original design, analytics counter, and links.
Adding a short description section.
Implementing two galleries (art by artists and AI) with dynamic random image switching over time.
Implementing image enlargement for a more detailed view (with a caption containing the art title and artist name, along with a link to their DeviantArt page for the artist section) when clicked, with a pause while viewing the enlarged version.
Support for two languages with auto-detection and manual switching.
Implementing a GitHub pipeline for building and deploying to hosting.
Documenting all steps taken and instructions for manual actions with the website code (for example, starting a local server for debugging).
The AI handled the tasks I set for it with virtually no clarification. I can only recall two instances where I had to make significant adjustments:
When switching images over time, due to the nature of random number generation, it sometimes happened that the same adjacent images were updated. To fix this, I asked Claude to save the index history of the last four updated images and exclude these indexes from the random selection.
The gallery preview image in the AI-generated code used a fairly large original image, and in production, images were often loaded slowly instead of instantly. I asked Claude to generate and use 300x300 thumbnails for the gallery, and load the full image only on click. Here, the AI didn't generate code, but used ImageMagick and bash for batch processing of images.
That is, the clarifications are related to not entirely obvious aspects of implementation, and I don’t see anything wrong with the fact that the AI itself didn’t think of this.
In addition to programming tasks, the AI also came in handy for helping with problem solving. While the AI art has a uniform square format, the artists' art is represented by a whole "zoo" of aspect ratios, and the gallery thumbnail needs to be displayed in a unified format. To solve this issue, I simply asked Claude to analyze the resolutions of several dozen artists' artworks selected for the website and suggest a thumbnail format. The assistant handled this task admirably.
And this is what the club's website looks like now (also the logo was updated a little):
Art from artists
I mentioned AI more in the article, but the club began as a collection of art from artists, and that remains an integral part of it, so I can't help but include a few of my favorite pieces from the club's selection:
Origin of the name
My favorite sci-fi series is "Star Trek: Enterprise" about Earth's first starship with a Warp-5 engine, the NX-01 Enterprise. The club was named after this beauty.
And in general, art based on the Star Trek universe occupies a significant part of the selection.
Club symbol
I created the logo myself in LibreOffice Draw; its only connection to the name is the stylized Starfleet Command symbol in the center. It's more of a symbol of the convergence of various technologies on the path to an ever-expanding exploration of the universe, deep and far, for the benefit of humanity.
Here is a version of the logo refined with the help of AI:
Plans
The club's future development includes implementing a parser for online news sites about breakthrough scientific and technological achievements (a traditional algorithmic task), selecting the most interesting ones (planned to be delegated to AI), and publishing them. The core focus — art — is planned to remain the top priority, with news being a supplement, perhaps once a week.
Here are examples of sites planned for use as sources, based on Gemini's recommendations:
SciTechDaily: A site dedicated to news and articles in the fields of science, space, and technology from various universities and research organizations.
Live Science: Covers a wide range of topics, including space, health, the environment, and history, with an emphasis on making science accessible and answering curious questions.
ScienceAlert: Covers major breakthroughs and surprising discoveries in an engaging and easy-to-read format.
You can subscribe to an email newsletter on each of these sites.
Conclusion
Sci-Fi Club NX-01 is a live example of how modern AI technologies can be used to create and share an inspiring vision of the future. By combining the best works of artists and the capabilities of AI, we are building a platform that reminds us of the potential of humanity and the universe. My wife helps me run the club, so we've already formed a small team. Join us and look to a positive future together!
I'd love to welcome new subscribers! And, of course, the club is open to collaboration.
Live long and prosper 🖖
