Andrew Murphy Says Faster Code Isn't Bottleneck

This is a solid theory-of-constraints take for the AI coding era: faster code generation often just exposes bad system design faster. If your org cannot absorb more output, copilots become a queue amplifier, not a productivity win.

• –The article’s strongest point is that throughput is set by the slowest step, not the fastest one.
• –It connects AI-assisted coding to familiar failure modes: bloated PR queues, flaky CI, manual approvals, and slow deploy paths.
• –The piece also lands on the human bottlenecks that matter most: unclear product direction, decision latency, and weak post-launch feedback.
• –For teams adopting AI tools, the actionable metric is cycle time from idea to user value, not lines of code or PR count.