The slop cannons in your engineering org

A critical look at 'slop cannons' — engineers and designers who over-rely on AI agents, generating high volumes of low-quality output. Backed by data: GitHub AI-agent PRs grew 325% in six months with a 90% noise rate, CodeRabbit found AI-coauthored PRs have 1.7x more issues, and a METR study showed AI made experienced developers 19% slower despite them feeling 20% faster. The post covers how LLM sycophancy enables this behavior, provides a manager's checklist to identify slop cannons via revert rates and PR size, and offers actionable fixes including capping parallel agents, mandating specs before agent runs, and forcing adversarial review. Also warns about cognitive offloading eroding junior developers' foundational skills.