Podcast
Scott Hanselman and Mark Russinovich on AI, engineering craft, cloud, career growth, and whatever technical rabbit hole is worth following this week.
Double-Check References
June 3, 2026
After scanning thousands of conference submissions, Mark finds hallucinated and inaccurate citations almost everywhere. He and Scott use it as a way into a bigger argument about accountability, taste, and where leaning on AI quietly erodes the judgment it is supposed to support.
Recent episodes
Have Taste
Where does product taste come from? Scott and Mark make the case that it is built from years of exposure and pattern recognition, not handed out as innate talent, and talk about how to develop it deliberately.
ZoomIt, Evolved
A build session on ZoomIt’s new panoramic capture and webcam overlay. Mark gets into stitching frames across any application, the edge cases that break it, and how AI-assisted coding sped the work up without removing the hard parts.
Sculpt, not Spec
Mark and Scott push back on the idea that everything starts with a finished spec. The work they describe looks more like sculpting: build a rough version, react to it, and let human judgment steer the AI-assisted iterations.
Beyond the Vibes: How Models Learn and Stitch Panoramas
What is actually happening beneath the hype: how training shapes behavior, where incentives leak into a model, and why even something as visual as panorama stitching turns into a hard systems problem at the edges.
Vibe Coding, for Real (Again)
Mark walks through using modern models to implement a shared-memory transport for gRPC across Go and .NET, including where AI helped and where engineering judgment still mattered.
A Public 1-1 for Software Engineering Preceptorship
A conversation about engineering mentorship, apprenticeship, and what actually helps new engineers build durable technical judgment.
More episodes:YouTube playlistAcastApple Podcasts