Published: 12 May 2026
Summary
Software is fundamentally built the same way, but each wave of abstraction — from programming languages to frameworks — has shifted what developers manage. Now, AI coding agents, while making it significantly easier to develop software and applications, add another layer of abstraction. Today, my “source code” looks different: markdown files instead of boilerplate, MCP integrations for AI agents instead of manual wiring of APIs into application code, prompt libraries and custom AI coding agent “skills” instead of language syntax. Software engineering leaders must treat these assets as systems that accrue technical debt and demand versioning, maintenance, and security discipline.
Included in Full Research