Agents Didn't Lower the Bar for Good Architecture. They Just Removed the Penalty for Ignoring It.

Boris Tane's piece on slop creep is worth five minutes of your time, but not for the reasons the headline suggests.

The argument most people will take away is "coding agents produce bad code." That's not the argument. The actual insight is buried in the middle, and it's more precise than anything else currently circulating on this topic.

Before agents, a developer with poor architectural instincts was rate-limited by their own speed. Bad abstractions compounded slowly. The pain surfaced in time to be fixed. Agents shatter that feedback loop — you can now ship a month's worth of structural mistakes in a week, each commit individually defensible, the damage invisible until it's load-bearing. The problem isn't bad output. It's plausible bad output, arriving faster than any review process was designed to catch.

That's the circuit breaker insight. It's the one thing in this piece that other AI productivity coverage consistently misses.

What the piece doesn't earn is the leap from "this happened to my side project" to "this is happening everywhere right now." The severity of slop creep scales sharply with how much review discipline a team actually has. A senior engineer reviewing PRs from a coding agent is still doing exactly what they were always doing: catching the decisions the implementer couldn't see. The circuit breaker isn't gone — it's just been pushed up a level.

Tane is describing a real failure mode for solo builders and under-reviewed teams. He's presenting it as an industry-wide crisis. Those are different problems with different solutions, and conflating them lets the actual audience — the person vibe-coding alone at midnight — off the hook for the specific choices that make this preventable.

The last paragraph half-acknowledges that the whole argument might not survive the next model release. After two thousand words building the case, that reads more like insurance than genuine uncertainty.

Read it for the circuit breaker argument. Be sceptical of everything after "the answer is not to stop using agents."