.NET Rocks! The Role of AI in Software Development
10 snips
Dec 18, 2025 Jeff Fritz, a Principal Program Manager on Microsoft's .NET community team, and Bill Wolf, a seasoned Philly.NET organizer, dive into the transformative impact of AI on software development. They tackle listener concerns about the costs and controls associated with AI technologies, discussing various approaches to tool utilization. The duo explores 'vibe coding' for rapid application prototyping, AI's role in specialized engineering, and the necessity for careful constraints around AI permissions. They emphasize embedding AI to enhance user experiences and predict a shift toward efficient, localized models post-bubble.
AI Snips
Chapters
Transcript
Episode notes
Monitor Model Usage And Choose Wisely
- Track and control model usage to avoid surprise bills by monitoring Copilot/Claude usage gauges in your editor.
- Choose lower-cost models for routine tasks and reserve expensive models for heavy code generation.
Use Strong System Prompts And Personas
- Constrain LLM behavior with a strong system prompt that forbids unintended actions and enforces style.
- Use curated persona and prompt templates (e.g., CoPilot Awesome repo) to get consistent, lower-cost outputs.
Define Agent Roles And Limit Permissions
- Build agents with clear roles (persona, scope, and allowed actions) and limit their permissions to avoid runaway behavior.
- Keep agents focused (tests, code, docs) rather than granting broad production-level access.