The Pluralsight Podcast · Tuesday · 1 hr 4 min

GitHub Unleashed: LLMs on Guard | Tim Warner

What happens when you don't just ask one LLM to check your work, but three, and then make them vote? In this episode of The Pluralsight Podcast, we sit down with Tim Warner, Principal Pluralsight Author, Microsoft MVP, and author of more than 200 Pluralsight courses, to talk about what it takes to stay current through three decades of tech disruption, and why gen AI is just the latest wave to ride rather than fear. Tim breaks down what technology and L&D leaders get wrong in the AI gold rush, why shadow AI is the new shadow IT, and how to pair training with guardrails and sandboxes your teams will actually use. Then Tim takes us inside GitHub for a live demo: a GitHub Actions workflow that hands agent and skill definitions to three different LLMs, which scan for prompt injection and other vulnerabilities, then vote on a verdict through a simple quorum mechanism. It's a practical look at defense in depth for AI-assisted development, explained in plain language. Chapters: 01:42 Eternally curious: 30 years in IT 05:07 Know yourself, do your best 07:11 The eternal student persona 09:04 Learning in public 10:57 Underestimated fundamentals and agentic AI risk 13:00 AI councils, not fear-based reactivity 15:24 Shadow AI is the new shadow IT 17:09 Closing the upskilling gap 19:41 The skills leaders underestimate 21:05 Prompt vs. context engineering 22:29 What GitHub actually is 26:17 GitHub Actions explained 29:32 Tim's multi-LLM research workflow 32:53 When LLMs disagree 35:15 Multi-model on a budget 38:30 Is on-prem AI coming back? 39:52 Getting started responsibly 43:38 Demo: LLM quorum security scanning in GitHub 53:29 Security risks and data handling 56:14 Enterprise guardrails and orchestration 59:47 Advice: curiosity over fear 1:02:06 Where to find Tim Connect with Tim Warner:TechTrainerTim LinkedIn: https://www.linkedin.com/in/timothywarner/ Website: https://techtrainertim.com Pluralsight courses: https://app.pluralsight.com/profile/author/tim-warner Stay up to date on everything happening in cloud, AI, and security. Sign up for our weekly newsletter: https://www.pluralsight.com/technews/

0:00-1:04:56

Play in

Transcript file

transcript

No transcript — this publisher did not publish one.

show notes

What happens when you don't just ask one LLM to check your work, but three, and then make them vote?

In this episode of The Pluralsight Podcast, we sit down with Tim Warner, Principal Pluralsight Author, Microsoft MVP, and author of more than 200 Pluralsight courses, to talk about what it takes to stay current through three decades of tech disruption, and why gen AI is just the latest wave to ride rather than fear. Tim breaks down what technology and L&D leaders get wrong in the AI gold rush, why shadow AI is the new shadow IT, and how to pair training with guardrails and sandboxes your teams will actually use.

Then Tim takes us inside GitHub for a live demo: a GitHub Actions workflow that hands agent and skill definitions to three different LLMs, which scan for prompt injection and other vulnerabilities, then vote on a verdict through a simple quorum mechanism. It's a practical look at defense in depth for AI-assisted development, explained in plain language. Chapters:

01:42 Eternally curious: 30 years in IT

05:07 Know yourself, do your best

07:11 The eternal student persona

09:04 Learning in public

10:57 Underestimated fundamentals and agentic AI risk

13:00 AI councils, not fear-based reactivity

15:24 Shadow AI is the new shadow IT

17:09 Closing the upskilling gap

19:41 The skills leaders underestimate

21:05 Prompt vs. context engineering

22:29 What GitHub actually is

26:17 GitHub Actions explained