In this episode of Scott & Mark Learn To, Scott Hanselman and Mark Russinovich discuss using AI to build a fractal shader in just 15 minutes—something that would normally take hours of manual coding. They explore the challenges and limitations of AI in real-world projects, particularly with large, complex codebases. The conversation covers the importance of understanding licensing and tools like GitHub Copilot for sourcing and attributing code. Mark shares his experiences with AI-assisted coding, particularly the frustrations of models that "fix" issues but miss the mark. They highlight how AI struggles with complex projects, maintaining state, and adapting to evolving requirements. Ultimately, they agree that while AI can assist with certain tasks, human expertise is essential for ensuring the right context and specification in development.

Takeaways:

• Despite initial skepticism, the first attempt at creating the shader worked nearly perfectly
• AI models may not always use the latest API versions, causing security or functionality issues
• Enterprise AI development involves continuous iteration, evolving from V1 to more refined versions

Who are they?

View Scott Hanselman on LinkedIn

View Mark Russinovich on LinkedIn

Watch Scott and Mark Learn on YouTube

Listen to other episodes at scottandmarklearn.to

Discover and follow other Microsoft podcasts at microsoft.com/podcasts

Download the Transcript

Hosted on Acast. See acast.com/privacy for more information.