Death Planning, Agent Battles, and Helicopters at MMS Midway
Andrew Pla sets up shop on the San Diego waterfront at MMS Midway Edition for a pair of bar sessions, with the occasional helicopter flyover from the USS Midway providing the soundtrack.
First up is Matt Zaske, longtime MMS regular and self-described conference mascot (stickers and all). Matt talks about his two sessions this year: a GitHub, CI/CD, and Actions deep dive aimed at sysadmins who aren’t developers, and a happy hour session on death planning and preserving your digital life. He makes the case that planning for the people you leave behind is as much about emotion as it is about tech, and that knowing what doesn’t matter (goodbye, home lab) is just as important as knowing what does (the Home Assistant setup your family depends on). Matt also shares how he’s using AI as a “trust but verify” sidekick, from writing ESP32 code he’d rather not fight with Python over, to having GitHub Copilot run code reviews on his own pull requests as a solo operator.
Then returning guest Danny Stutz, fresh off his first time speaking at MMS, breaks down his coding agent battle royale. Danny built an arena that runs the same prompt against multiple AI coding agents using the same model, then compares completion time, token usage, and cost. The audience-submitted experiments ranged from a protein powder price finder to a Tinder for lost socks (complete with a “match” button that runs away from your mouse). Danny shares why open source harnesses like Pi and Aider stood out, why a simple “hello” can burn thousands of tokens, and why AI spend tracking is about to become a big deal for companies. The conversation wraps with Danny reflecting on his first big career risk and a reminder that taking chances, making mistakes, and finding your people is how growth happens.
Key Takeaways:
Digital legacy planning matters for sysadmins. Think about who will actually be handling your stuff, what they can realistically execute in an emotional moment, and how your open source projects and home infrastructure live on after you.
AI coding agents differ more than you think. With the same prompt and the same model, different harnesses vary widely in speed, token usage, and cost, and hidden context like system prompts and safety instructions adds up fast.
Trust but verify, and keep putting yourself out there. Use AI to lower the barrier to entry, validate what it produces, and lean on community to push past imposter syndrome and take the next career risk.
Guest Bios:
Matt Zaske is a repeat MMS speaker and longtime community member, Matt is known for his sessions on task sequences, 3D printing, and Home Assistant, as well as his humor, storytelling, and the stickers/3D prints he hands out to help attendees break the ice.
Danny Stutz is a cybersecurity and enterprise technology architect at West Monroe, where he has worked for over seven years helping clients with AWS, Azure, Microsoft 365/Entra ID, and IAM services. A returning PowerShell Podcast guest and first-time MMS speaker, Danny focuses on AI, data science, and putting coding agents to the test.
Resource Links:
MMS Midway Edition schedule: https://mmsmidway.sched.com
Matt Zaske’s blog: https://www.mzonline.com
Matt Zaske on Sessionize: https://sessionize.com/zaskem
Danny Stutz on LinkedIn: https://www.linkedin.com/in/daniel-stutz/
Danny Stutz on X: https://twitter.com/danny_stutz
Pi coding agent: https://pi.dev
Aider: https://aider.chat
PDQ Connect PowerShell Scanner: https://www.pdq.com/blog/the-powershell-scanner-has-arrived-in-pdq-connect/
PDQ Discord: https://discord.gg/PDQ
Connect with Andrew Pla: https://andrewpla.tech/links
The PowerShell Podcast: https://www.pdq.com/resources/the-powershell-podcast/
The PowerShell Podcast on YouTube: https://youtu.be/ZIc_AStvaNo