「グラフエンジニアリング」って結局何？ - AIスロップによる汚染に注意 -The article traces the rapid succession of AI-development buzzwords—from prompt…

The speed at which the AI-development community invents new vocabulary has itself become a talking point. Over roughly the past year, the field has cycled through "prompt engineering," "context engineering," "harness engineering," and "loop engineering," and a new candidate has now surfaced: "graph engineering." A blog post published on Zenn traces this rapid succession of buzzwords and raises a skeptical question—whether "graph engineering" marks a genuine paradigm shift in how developers work with large language models, or whether it is closer to AI slop, a superficially convincing term produced and spread faster than anyone can assess its substance.

The concern matters because terminology shapes how teams allocate attention, tooling, and budget. If a phrase gains traction without a clear technical definition, developers may end up chasing a label rather than a practice.

The post frames the progression as a lineage. Prompt engineering, the earliest of the group, refers to carefully wording the input given to a model in order to steer its output. Context engineering broadened the focus from the single prompt to the full set of information placed in a model's context window—system instructions, retrieved documents, tool definitions, and conversation history. Harness engineering shifted attention to the scaffolding around the model, the code and infrastructure that feed it inputs and handle its outputs. Loop engineering reflected the rise of agentic systems, where a model iterates in a cycle of acting, observing, and revising rather than answering in a single pass. Graph engineering, the newest entry, appears to extend this idea toward representing workflows as connected nodes and edges.

According to the post, the immediate trigger was a message from Peter Steinberger, described as the developer of OpenClaw. The excerpt points to a post that provocatively asked, in effect, whether people were "still talking about loops, or have already shifted to graphs." The following day, the post notes, a roadmap-style article appeared—an example of how quickly a casual remark can harden into what looks like an established direction.

This is where the "AI slop" warning enters. The term "AI slop" generally describes low-quality, mass-produced content generated by AI systems and pushed into circulation with little human vetting. The post's argument is that the "graph engineering" label may be an instance of this phenomenon: a term that sounds like a natural next step in an evolving series, and is therefore easy for both people and models to generate, repeat, and treat as real, even if it lacks a rigorous definition or demonstrated practice behind it.

There is