DX studied more than 500 engineering organizations using AI. Pull requests got bigger, reviews got slower, and the steady, incremental delivery went down. Maintainability scores rose while confidence in each change fell.
Writing got cheap. Judging did not. When one side of the work speeds up and the other stays the same, the slow side becomes the whole system. Nobody slowed down on purpose; the checking could not keep pace with the making.
Your speed is set by the part you cannot hand off: knowing whether what arrived is right.
Kay
P.S. Course 0 is where these ideas become practice: https://systemthinkinglab.ai/course-0 This note goes out by email Sunday to Friday. Read more at systemthinkinglab.ai.
source & further reading
dev.to — original article
Aimeos 2026.10: MCP tools for Laravel e-commerce
How Loopers-OSS Saves 76% of Input Tokens and 68% of LLM Spend
No MCP? No problem. How I got Instinct AI to call a flight API