Demystifying AI terms: loop engineering, squads, and harness | S02E02 | The GitHub Podcast

GitHub’s podcast episode with Cassidy, Marlene, and GPS breaks down emerging AI terms—loop engineering, harness engineering, and squads—and connects them to practical agentic workflows, including how iteration loops affect cost and performance and how “open weight” differs from truly open source AI.

Overview

In this episode of The GitHub Podcast (S02E02), the hosts unpack fast-evolving AI/agentic-workflow terminology and how these concepts show up in real-world agent systems.

Topics covered

• Navigating confusing AI lingo
• Loop engineering
  • What “loop engineering” means in the context of agentic workflows
  • Discussion of Ralph loops and how loops relate to costs
• Squads
  • The idea of parallel agent collaboration (multiple agents working concurrently)
• Harness engineering
  • What a “harness” is and how it’s used to run/evaluate agent workflows
• Hill climbing
  • Using iterative improvement techniques to improve agent performance
• Forward deployed engineer
  • Discussion of the role in the context of applying these systems in practice
• Model openness terminology
  • Differences between closed models, open-weight models, and true open source AI

Open source picks mentioned

• Astro v7
• RAE API
• Hugging Face Tao