- Teenage builder and vibecoder fascinated by agentic AI.
- I see AI as a way for people without traditional coding experience to turn strong ideas into working products.
- Interested in AI-agent architecture—especially how one person can coordinate specialised agents to research, reason, and act.
- Constantly asking: How can we trust agents as they gain more autonomy and access to personal data?
- Created Critiqor to observe agent behaviour, diagnose runtime risks, and improve reliability through evidence.
- I build open-source tools at the edge of my understanding to solve real problems and make my learning visible.
- My goal is to make AI agents more accurate, reliable, and understandable for the people who depend on them.
🌐 Visit my website: web3curtis.vercel.app
Runtime intelligence for AI agents.
Critiqor helps developers observe agent execution, diagnose reliability risks, and improve agent behaviour using evidence from the runtime—not only the final response.
View repository → · Explore Critiqor →
An experiment in making agent–website interactions more inspectable and reliable.
Built for the WebMCP Challenge, this project demonstrates how runtime evidence and authoritative state reconciliation can prevent duplicate effects after an agent loses a tool response.
View Devpost submission → · Try the live experiment →
- AI-agent architecture
- Runtime observability and evaluation
- Agent reliability and accuracy
- Human–agent collaboration
- Open-source developer infrastructure
