.NET Rocks! Building an AI App with Calum Simpson
10 snips
Dec 4, 2025 Calum Simpson, a solution architect and VP of AI at SSW, dives into the innovative world of AI-centric applications with a focus on his product, YakShaver. He explains how YakShaver streamlines issue reporting, emphasizing the importance of using LLMs and the upcoming V2's integration with Model Context Protocol servers. Calum also shares insights on balancing flexibility and safety in AI tools, the concept of vibe coding for development, and the potential of AI to enhance project delivery by reducing busywork and fostering creativity.
AI Snips
Chapters
Transcript
Episode notes
Automate Reporting By Capturing Context
- Yak Shaver targets the repeated busywork around reporting issues by automating context capture and routing.
- Using video input plus LLMs lets teams record problems once and create consistent backlog items automatically.
Constrain LLMs With Pipelines Then MCP
- Guide LLMs through constrained pipelines rather than granting full autonomy to reduce unpredictable behavior.
- Move to MCP servers later to keep flexibility while preserving safe, auditable tool actions.
Video Input Preserves Repro Steps
- Video-plus-transcript input gives richer, replayable context than text-only reports and reduces back-and-forth repro steps.
- Embedding the original video link in the generated PBI lets developers ignore AI prose and view the reporter's demonstration.