Software engineering has always been context engineering¶
Insight: Context engineering isn't a new discipline created by AI — it's what software engineering has always been. At Google, engineers spent significant time using Code Search to gather context, following code paths, adding debugging statements, even printing code on paper to understand it. The core skill of gathering, organizing, and applying the right context to make decisions has remained constant across a decade of tech evolution; AI tools just made the pattern explicit.
Detail: YK (who built the open-source AI coding agent Kaguya and introduced agentic coding to Sourcegraph's Cody) distills context engineering for AI to three principles: (1) make instructions as clear as possible with lots of detail, (2) provide enough relevant context so the agent doesn't guess, (3) keep context as short as possible for focus. The parallel with traditional SWE is deliberate — using AGENTS.md is like writing onboarding docs, using MCP is like setting up Code Search, using focused prompts is like scoping work tickets.
Sources
Related: existing entry "Context engineering supersedes prompt engineering" in batch-1/claude-code.md — CORROBORATES