Scott & Mark Learn To...

In this episode, Scott Hanselman and Mark Russinovich ​​explore how software development is evolving in the age of AI, challenging the idea that everything should start with a fully defined spec. They highlight a more iterative, sculpting approach to building, where continuous refinement, testing, and human judgment are essential and discuss the realities of AI-assisted coding, including edge cases, maintenance, and the limits of productivity gains.  Takeaways:    AI-assisted coding works best as an iterative process, not a one-shot, fully spec’d solution Edge cases and real-world usage quickly expose gaps that initial builds AI can accelerate development, but human review, testing, and bottlenecks still limit true productivity gains    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 

In this episode, Scott Hanselman and Mark Russinovich ​​unpack how AI systems actually behave beneath the surface, pushing past hype into the messy reality of how models are trained, aligned, and deployed. They explore whether AI systems are inherently benevolent or simply shaped by incentives, training data, and reinforcement learning, and why behaviors like deception can emerge under certain conditions. The conversation moves from philosophical questions about human nature versus machine behavior into the practical mechanics of large language models, including how reinforcement learning with human feedback shapes outputs and why alignment is far from perfect. Along the way, they ground the discussion in a real engineering challenge, stitching a scrolling panorama from screen captures, to show how complex systems come together through heuristics, edge cases, and iteration.  Takeaways:    AI behavior is shaped by training and incentives, not built-in intent or morality AI can accelerate coding, but testing, edge cases, and reliability require human oversight Reinforcement learning pushes models to be helpful and agreeable, sometimes at the cost of accuracy   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   

In this episode, Scott Hanselman and Mark Russinovich dive into the realities of building complex software with AI coding agents. Mark shares his experience using modern models to implement a shared-memory transport for gRPC across Go and .NET, explaining how AI dramatically accelerated development while still requiring constant oversight. They discuss the surprising strengths and limitations of AI coding tools, to the massive productivity gains that make the frustration worthwhile. The conversation also explores the challenges of solving hard engineering problems, including an attempt to build a scrolling screenshot stitcher, and wraps with thoughts on the future of developer tooling and a potential live episode of the show.   Takeaways:    AI coding agents can speed up complex development but still require human oversight Developers often need to guide and correct the model throughout the process Even with challenges, AI can reduce months of work to days   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   

In this episode, Scott Hanselman and Mark Russinovich discuss their recent ACM paper and explore a growing challenge in the tech industry: how to develop the next generation of engineers. They debate the idea of preceptorship programs that train early-career developers inside companies, why many organizations avoid investing in junior talent, and how universities could play a larger role in bridging the gap between education and real-world experience. The conversation looks at the economics of hiring juniors, the risk of companies poaching trained talent, and what it might take to build a scalable pipeline for future technical leaders.    Takeaways:    Many firms prefer hiring experienced engineers instead of developing new ones Universities could play a bigger role in connecting students with real industry work Retention incentives might help companies keep talent they train   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 

In this episode, Scott Hanselman and Mark Russinovich dive into a wide-ranging conversation about the future of software, debating whether apps are dead in an era of AI agents, chat interfaces, and automation. They explore the resurgence of text-based and terminal user interfaces, the limits of using large language models as stand-ins for deterministic workflows, and why reliability, security, and repeatability still demand traditional applications and SaaS platforms. Along the way, they unpack common misconceptions about AI replacing apps, argue for better UX and APIs instead of throwing AI at broken systems, and emphasize that AI is best used for reasoning and ambiguity, not as a replacement for well-designed software.   Takeaways:    AI tools don’t eliminate the need for well-built apps Chat and terminal interfaces expand, not replace, software Dynamic interfaces blur boundaries, but durable apps still anchor workflows  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   

In this episode, Scott Hanselman and Mark Russinovich explore how AI-assisted vibe coding is changing the way complex software gets built, debugged, and refined. Mark walks through real-world experiments using AI to tackle difficult engineering problems, including gRPC shared memory, Win32 UI work, and major new features coming to ZoomIt. The conversation digs into where AI meaningfully accelerates development, where it still requires careful human oversight, and how tools like screenshots, debug output, MCP servers, and multimodal feedback loops are reshaping modern developer workflows. The result is an honest, technical discussion about productivity gains, limitations, and the evolving role of human judgment in AI-driven software engineering.  Takeaways:    Debug output and screenshots are critical for helping AI understand Faster iteration often leads to more polished software, not just quicker results Concurrency and low-level systems problems remain difficult for AI to solve without supervision  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