Leading Humans and Agents: Hybrid Teams as Complex Adaptive Systems with Diana Larsen

In this Mob Mentality Show episode, Chris Lucian sits down with Diana Larsen — co-author of Lead Without Blame, Five Rules for Accelerated Learning, and Agile Retrospectives — to explore what leadership looks like when your team is made of both humans and AI agents. Diana, who came into agile "through the XP door" in the late 90s, argues that a team of some number of humans plus some number of agents is still a complex adaptive system: it has a container, an identity, differences among its players, and exchanges that let it collaborate. The practices we built for human teams — retrospectives, chartering, liftoffs, checkpoints — still apply, if we're intentional about bringing them along.

Diana shares her "your agent is your mirror" idea (inspired by the book Your Dog Is Your Mirror): LLMs reflect the behavior they were trained on, and the way we talk to them shapes us too. She types please and thank you to her agents — not for them, but to retain her own humanity. Chris connects this to Theory X vs. Theory Y prompting, his agents.md files with single and double loop learning built in, and his experiments using a frontier model to build a deterministic TDD state-machine MCP server that keeps a less capable model honest — like a senior developer building a process for a junior.

The conversation then turns to mobbing with agents and the leadership challenges of this moment. Chris describes how his mobs use "weighing the ox" (wisdom of the crowd) prompting, collaborative prompting, and rotating across four machines, and why the 20-minute agent runtime is a key threshold for context switching. He argues mobbing is more valuable than ever because the typing was never the hard part. Together they discuss the diffusion of innovation, why "Agile is dead" gets declared at every chasm crossing, token maxing as Goodhart's Law taken to its extreme, the Red Queen hypothesis, why no one should have been laid off over AI, and a broken hiring pipeline where a "beautiful" new-grad resume from three years ago now gets 6 callbacks out of 700 applications.

We dig into:

• Human + agent teams as complex adaptive systems — and why the old team practices still apply.
• "Your agent is your mirror": blaming language, hallucinations, and retaining our humanity by saying please and thank you.
• Theory X vs. Theory Y prompting, and agents.md files that invite single and double loop learning.
• Why you should value stream map and remove waste before going agentic — or your agents will just perform bureaucracy.
• Keeping prompts small and bite-sized (from Five Rules for Accelerated Learning) while still leaving room for emergence.
• "AI gaslighting" in TDD and using a frontier model to build a deterministic MCP server that a smaller model must follow.
• Mutation testing and building friction points into flow systems to check you're still heading the right way.
• The naive observer in the ensemble — Diana's Roman numerals kata story from one of Woody Zuill's workshops.
• "Weighing the ox" prompting, collaborative prompting, and the 20-minute threshold for short vs. long agent workflows.
• Why mobbing is more valuable than ever when the typist role has become minimal.
• Leading through the diffusion of innovation: AI-forward vs. AI-resistant people, token maxing, and Goodhart's Law.
• The Red Queen hypothesis and why laying off developers over AI was a bogus move.
• A broken pipeline into jobs: resumes, interviews, and university curricula that need to change.
• Agile is dead (again?), the iPod wheel, and how innovations reshape as they cross the chasm.
• Why AI is unwittingly nudging people into XP, lean, and good work process — tests included.

Podcast Episode… Apple: https://podcasts.apple.com/us/podcast/the-mob-mentality-show/id1485950034 Spotify: https://open.spotify.com/show/01npjdd1VRiNqw1RZ8aB5g?si=607de4da208a425c PodBean: https://mobmentalityshow.podbean.com/ Amazon: https://music.amazon.com/podcasts/7a695161-1857-466d-83ca-d9a75161823b/the-mob-mentality-show

References… Diana Larsen's LinkedIn: https://www.linkedin.com/in/dianalarsenagileswd/ Diana Larsen's website (Thrive in Turbulent Times program): https://dianalarsen.com/ Lead Without Blame by Diana Larsen and Tricia Broderick Five Rules for Accelerated Learning by Willem Larsen and Diana Larsen Clickchain.ai: https://clickchain.ai/ Your Dog Is Your Mirror by Kevin Behan Diffusion of Innovations by Everett Rogers Pacific Northwest Software Quality Conference: https://www.pnsqc.org/ OOP Conference Munich: https://www.oop-konferenz.de/

Thanks to G-SLiK (https://soundcloud.com/g-slik) for the intro and outro music.

Chris Lucian and Austin Chadwick discuss all things #agile and product development from a #MobProgramming perspective.

Chris Lucian is the Director of Software Development at Hunter Industries and a founder of mob programming. https://www.chrislucian.com/p/chris-lucian-biography.html

Austin Chadwick is a Mob Programmer at Hunter Industries and is a passionate agilist and craftsman with experience in several roles (e.g. coach, developer, tester, scrum master, business analyst). https://www.linkedin.com/in/austin-chadwick-3a58151a4/

We would love your feedback and ideas for future episodes! Please add comments to the video or reach out to us on Twitter ( https://twitter.com/mob__mentality & https://twitter.com/ChristophLucian ).

All statements and opinions expressed by Chris, Diana, and Austin are solely their own and do not represent the views of any company. Chris and Austin are just sharing and not recommending ( https://justsharing.dev/ ).