I recently challenged myself to build an AI agent that could answer questions about me as both a designer and a person. The result? A functional prototype completed in just 5 hours.
How I Built It
Using ChatGPT, OpenAI’s API, Netlify for hosting, and GitHub for version control, I created an agent that pulls from markdown context files to provide personalized responses. The agent is integrated with my existing WordPress portfolio site at ryan.green/ai-agent.
Key Learnings:
- Chunking beats delegation: Building in smaller chunks with LLM guidance produces better results than handing over 100% control to AI tools like Cursor. One-shot solutions aren’t quite there yet.
- Context is king: GPT-5’s expanded context windows make troubleshooting remarkably efficient, helping resolve issues from server logs almost instantly.
- Ship first, polish later: For smaller projects, I’ve found the design thinking process has inverted. Getting a working prototype out first and then refining the design leads to much faster shipping than traditional design-first approaches.
Next up: I’ll be expanding the agent’s knowledge by adding context from my Notion workspace and giving the interface some visual polish.
This experience reinforced a powerful insight for designers today: start with outcomes, build something functional, then inject taste. That’s how designers become builders in today’s fast-paced environment.