AI Co-Piloting
Architected entirely through a conversational pipeline, utilizing large language models for both logic scaffolding and structural generation.
Win Twin
An innovative collaboration project between AI and John G, exploring the intersection of human creativity and artificial intelligence.
Visit Live SiteTechnical Architecture
Rapid Iteration CI/CD
Setup automated deployment hooks to allow the AI-generated code to be pushed to a live testing environment seamlessly.
Prompt-Driven CSS
The visual language was established through descriptive natural language prompts translated directly into raw CSS.
Obstacles & Solutions
Obstacle 1: Hallucinated Code Errors
The Issue: Because large portions of the site were generated by LLMs, there were frequent instances where the AI would hallucinate unsupported APIs or generate broken DOM structures.
The Solution: I built a rigorous human-in-the-loop validation step. Rather than accepting the code wholesale, I treated the AI as a junior developer, breaking down tasks into extreme micro-components to isolate errors before integration.
Obstacle 2: Context Window Limits
The Issue: As the project grew, the AI would "forget" the established CSS variables and design systems created earlier in the session.
The Solution: I authored a central "knowledge base" markdown file that was fed into the prompt at the start of every new session, ensuring the AI strictly adhered to the design rules regardless of how long the conversation went on.