在 QCon 中创建有效学习体验的一种由 AI 驱动的方法

Main speaker and conference details: Wes Reisz, technical principal and conference chair for upcoming [QCon.Ai], spoke at [InfoQ Dev Summit Boston] about a program led using AI during QCon London 2025.
Takeaways:
- Build the Right Thing: Combine lean thinking with AI for validating understanding and iterating towards MVP.
- Embrace Change: Keep learning as AI innovation is fast.
- Human Connection Remains Essential: People value human interaction even with AI tools.
Certification program experiment at QCon London 2025: A certification program was made possible with AI. It started with low-fidelity experiments, limited to 30 in the first cohort and expanded at QCon San Francisco. It included special events during the conference and an AI-driven workshop.
Workshop details: Used Retrieval Augmented Generation (RAG) with GPT-4 Turbo and external data sources. Pulled 75 conference talk videos into a vector database. Attendees could ask targeted questions using mobile devices.
System development: Designed architecture on a whiteboard and iterated with an LLM. Used Supervised Agents to break specification into features. 95% of code generated using Cursor and Claude 3.7 Sonnet, and wrote 5% code manually.
Workshop process: Started with a one-hour interactive discussion, then open space for attendees to discuss ideas and pick topics. Ended with wrap-ups.
Interview responses:
- Used Cursor and Claude Sonnet 3.7 with one-shot prompting and metaprompting. AWS services for transcription and storage. GitHub Actions and Terraform for CI/CD and IaC.
- Focused on functionality with tests. Each feature was reviewed via PR process before merging.
- Main lessons: Build the right thing with lean thinking. Generated code quality needs improvement but is only going up. Embrace change as AI is changing system creation.