# AI in Practice: Generating Art, Ideas, and Code for a Sci-Fi Project
A software engineer created the online club NX-01, dedicated to positive sci-fi art. The project combines manual curation of artists' works with AI-generated images, ideas, and code. Publications go out twice a day: one from artists, the second—AI art based on specific prompts. The audience exceeds 4,000 on VKontakte, plus Fediverse and Telegram. It uses Imagine Art (ex-Midjourney), Gemini, and Cursor with Claude 4.5 Opus.
Generating Sci-Fi Art
The "Future People" series filled a gap in depictions of peaceful characters. Prompts follow this structure: core idea + descriptors (age, ethnicity) + positive sci-fi style + details. Example: "adult experienced kind sci fi asian woman engineer".
AI often ignores excessive details, so prompts are kept simple. Adding "positive" is crucial: without it, the future looks grim; with it, it's bright and optimistic. Art themes:
- Space exploration: minimalist starships made of chrome rings with blue beams, astronauts in open space.
- Robots in peaceful scenarios: playing basketball in a hard sci-fi style.
- Futuristic cities: solarpunk metropolises with vertical gardens and waterfalls.
- Nanotechnology: nanobots repairing cells.
- Time travel: a portal in a lab overlooking a prehistoric world.
- Everyday life in the future: a classroom with a teacher and students.
Prompt structure:
- General description (woman engineer).
- Descriptors (adult, experienced, Asian).
- Positive vibe.
- Details as needed.
Quality is high, with artifacts being rare.
Generating Ideas for Art
Gemini is used for daily content. AI suggests concepts like "futuristic colony in a white quartz canyon on a planet with red sands." Implementing them in Imagine Art yields detailed scenes.
Hallucinations happen: "gothic cathedral-ship made of white stone." Such ideas are filtered out—only realistic ones get published.
Automating the Website with AI
The club's site migrated from HTML/CSS to React without any manual coding. Cursor with Claude handled:
- Dynamic galleries for artists and AI art with a rotation timer.
- Modal image zoom with captions and links.
- Bilingual support (auto-detection + toggle).
- GitHub Actions for CI/CD.
- 300x300px thumbnails (generated via ImageMagick/bash).
- History of recent indexes to avoid duplicates in rotation.
- Documentation and local setup instructions.
Clarifications were needed twice: for rotation without duplicates and loading optimization. Claude analyzed artists' art resolutions and recommended thumbnail formats.
Frontend tasks were solved by the assistant, despite the creator's focus on Linux/Python/QA.
What Matters
- Art prompts: Keep them simple + "positive sci-fi" for optimistic results.
- Ideas from Gemini: Solid concepts, but filter out hallucinations.
- Code from Claude: Full stack from React to CI/CD with no hands-on work.
- Optimizations: Thumbnails, rotation history, resolution analysis.
- Scale: 4k+ subscribers, daily posts.
— Editorial Team
No comments yet.