Neural Flux

• 

  Homepage

• 

Inspiration

I was inspired by the increasing reliance on large language models and the unseen complexity behind real-time AI systems. While AI tools appear effortless on the surface, they constantly process ambiguous, high-stakes, and emotionally varied requests under strict latency constraints. I wanted to explore what that experience would feel like if a human were placed in the role of the AI itself. Neural Flux was designed to simulate cognitive overload, prioritization, and response quality under pressure, turning abstract system behavior into an interactive game.

What It Does

Neural Flux is a real-time, pressure-based simulation game where I act as an artificial intelligence responding to a continuous stream of human-generated requests. Each request is dynamically generated and varies in complexity, intent, and emotional tone. The game challenges the player to balance response speed, semantic relevance, and helpfulness as multiple tasks compete for attention. Performance is evaluated using similarity scoring against AI-generated reference responses rather than exact string matching, encouraging meaningful, human-like answers.

How I Built It

I built Neural Flux using Python as the core language, structuring the project around modular game logic and AI interaction layers. The system integrates an AI model to generate both user scenarios and high-quality reference responses in real time. Player input is evaluated using semantic comparison techniques to assess relevance, completeness, and tone. I implemented progressive difficulty scaling by increasing prompt complexity, reducing response time windows, and introducing multi-part or emotionally nuanced requests as the game advances.

Challenges I Ran Into

One of the primary challenges was managing the trade-off between difficulty and playability. Increasing cognitive load too aggressively made the experience frustrating, while too little pressure reduced engagement. Another challenge was ensuring AI-generated scenarios felt realistic and diverse without becoming repetitive or unfair. Integrating AI responses in a way that supported gameplay—without introducing latency or unpredictability—also required careful prompt design and system constraints.

Accomplishments That I’m Proud Of

I’m proud of building a complete, solo-developed interactive system that meaningfully integrates AI beyond simple text generation. The game produces unique scenarios on every playthrough while maintaining consistent scoring logic and user feedback. I’m also proud of the project’s ability to communicate a broader concept—the operational pressure of AI systems—through gameplay rather than explanation.

What I Learned

Through this project, I gained hands-on experience designing real-time interactive systems, structuring AI-driven workflows, and evaluating natural language input beyond keyword matching. I learned how to control AI variability through prompt engineering, how to balance system complexity under time constraints, and how to build a polished solo project with clear architectural boundaries.

What’s Next for Neural Flux

Next, I plan to expand the semantic evaluation system to include finer-grained feedback on tone and intent, add adaptive difficulty based on player performance, and introduce visual and audio feedback that reflects system stress. I also want to explore deploying the AI logic as a lightweight backend service to support multiplayer or competitive modes.

Built With

• audio
• autoprefixer
• html2canvas
• javascript
• openai
• postcss
• react
• tailwind
• typescript
• vite