DevOps Co-Creator Urges Engineers: Build AI Systems, Not Prompts

Patrick Debois, who coined the term "DevOps" in 2009 and co-authored "The DevOps Handbook," has a clear message for software engineers: stop manually correcting AI-generated code and stop obsessing over prompts. Build the system instead.

Speaking at PlatformCon and other industry events, Debois argues that the software development lifecycle is evolving into what he calls a "context development lifecycle" (CDLC). This marks a shift from deterministic coding to managing the probabilistic systems that power AI agents.

The Context Development Lifecycle

The CDLC consists of four concentric loops that replace the traditional single CI/CD feedback loop: generate, evaluate, distribute, and observe. Engineers generate context, evaluate it through test-driven development and automated evals, distribute context as a packaged artifact, and continuously observe agent behavior in production.

"The mindset change for a developer that I would advocate is that if the agent didn't do what you want, stop correcting the code," Debois said. "Improve the system, not the prompt."

Harness Engineering

At the core of this approach is "harness engineering"—building the infrastructure, guardrails, and observability tools that wrap around large language models. This includes logs, metrics, and traces that allow AI agents to see the consequences of their actions and improve autonomously.

Effective agent context draws from diverse organizational knowledge: the codebase itself, specifications, post-mortems, documentation, Slack messages, architectural decision records, code samples, and compliance guidelines. The quality of this environment, not the LLM's raw capability, becomes the bottleneck for AI productivity.

ROI Depends on System Design

Debois warned that traditional feedback loops are insufficient for AI agent fleets. "Relying solely on traditional feedback loops means your AI agent fleet will not return on its investment," he stated at PlatformCon.

This transition redefines the value proposition of human engineers. Rather than tactical implementation through manual coding, engineers shift toward strategic system design and organizational knowledge management. The tension between AI efficiency and fears of developer layoffs dissolves when the engineer's role evolves from writing code to architecting the systems in which AI agents operate reliably and autonomously.

"The fundamental change is the system, not the prompt," Debois emphasized.

For an industry still grappling with how to integrate AI tools into existing workflows, Debois's framework offers a concrete path forward: invest in context quality and harness infrastructure, and let agents handle the implementation details.