Scott & Mark Learn To...

In this episode, Scott Hanselman and Mark Russinovich explore new AI security research revealing how smaller, unrestricted models can leverage more powerful AI systems to complete tasks those systems would normally refuse. Mark breaks down the research, its surprising results, and the potential security risks of AI models working together without understanding the bigger picture. Plus, Scott shares his approach to managing multiple AI agents, and the two discuss the challenges of keeping AI-powered development secure and efficient.



Takeaways:

Why breaking complex tasks into smaller pieces creates new AI security risks

How Scott uses AI agents to manage development tasks and reduce workload

What Mark’s latest research reveals about vulnerabilities in AI systems



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 try something new as Mark reacts to TikTok videos exploring the decline of reading, shrinking attention spans, and the impact of AI on critical thinking. What starts as a lighthearted experiment turns into a deeper debate about technology, education, and what it means to be human in an AI-driven world.



Takeaways:

How social media is changing our attention spans and reading habits

Why critical thinking matters in an AI-driven world

The risks of relying too heavily on AI to think for us



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 Mark’s new SemanticStart project, exploring how semantic search can make finding Windows apps, tools, and settings more intuitive. Along the way, they experiment with AI-assisted coding, discuss how agents could interact with local tools, and debate a bigger question: does prompting an AI actually count as programming?



Takeaways:

How semantic search combines keywords, embeddings, and enriched data

Why AI is changing the way developers build and prototype software

Whether prompting an AI should actually be considered programming.



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 security risks surrounding increasingly capable open-weight AI models and Mark’s latest research into a new defense called Fool’s Gold. Mark explains how safety guardrails can be removed from open models and explores an alternative approach: training models to provide convincing but deliberately flawed information when those protections are bypassed. They discuss how the technique works, whether it affects legitimate model behavior, and what it could mean for the future of AI safety as open models continue to advance.

Takeaways:

Why open-weight AI models can create new security risks

How Mark’s Fool’s Gold approach uses decoy responses as a defense decoy responses as a defense

Techniques that can protect against harmful use without degrading normal model performance

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 when to let AI solve a problem and when figuring it out is half the fun. They celebrate 30 years of Sysinternals by looking back at how the tools have evolved and how AI is changing software engineering and troubleshooting. Along the way, they debate efficiency versus learning by doing and consider a future where tools may increasingly be built for AI rather than humans.



Takeaways:

How Sysinternals has evolved over the past 30 years

How AI is changing software engineering and troubleshooting

Why future developer tools may be designed for AI, not humans



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

Produced by Hangar Studios