Matt Pocock Tames AI-Slopped Codebases

This is less about a single launch and more about codifying a workflow for surviving AI-generated mess.

• –The core idea is sound: AI increases throughput, but without discipline it also increases the rate at which bad code accumulates.
• –The post fits the broader “tracer bullets / small slices / fast feedback” philosophy, which is one of the few reliable ways to keep agent output reviewable.
• –The strongest value here is operational, not conceptual: it encourages teams to encode quality expectations into reusable skills instead of relying on memory or prompt folklore.
• –It will resonate most with engineers already using coding agents and feeling the maintenance cost rise faster than output.