Summary
IndyDevDan — software engineer Dan Eisler, with 15 years of experience spanning finance systems, game development, and early AI coding tools — takes direct aim at the “loop engineering” concept gaining traction in AI circles, arguing it is a confusing rebrand of the software development lifecycle that obscures more than it clarifies. His alternative framing: think in terms of AI developer workflows inside a software factory, with three distinct actors of value creation — engineers, agents, and code — each with different cost profiles, reliability characteristics, and appropriate roles.
The video walks through the actor hierarchy methodically: code is the fastest, cheapest, and most reliable; engineers are costly but provide judgment; agents sit between them, useful but expensive in tokens and less consistent than deterministic code. The power of agentic engineering, Eisler argues, lies in knowing when to deploy each actor and how to compose them into workflows that generate consistent business value.
He then builds up increasingly complex workflow patterns on screen: from a simple prompt-review loop to a full CI/CD-integrated build pipeline where a build agent generates code, a verification agent tests it in a sandbox, and human review gates the final ship. He also covers a hotfix scenario — production down — where a specialized surgical hotfix agent races against parallel sandboxes to find a solution fastest, with human-in-the-loop approval before deployment. Throughout, Eisler emphasizes that naming this correctly matters: clarity about what you’re actually building determines how well engineering teams can design, audit, and scale these systems. A rigorous, opinionated framework piece from a practitioner with receipts.
📺 Source: IndyDevDan · Published July 13, 2026
🏷️ Format: Opinion Editorial
