In a post on its own engineering blog, Anthropic says its continuous-integration workload grew 25-fold in six months as coding agents took over more of the work. Engineers shipped roughly eight times more code per quarter than they did between 2021 and 2025, with Claude authoring about 80% of it, and test volume rose tenfold. The company's response was test-impact analysis: selecting which checks to run based on what actually changed, rather than running everything. These are self-reported internal numbers, so treat the multiples as directional. Still, it surfaces a second-order cost of agentic coding that teams adopting these tools tend to discover late — if an agent writes code eight times faster, the verification infrastructure has to scale with it, and that bill is real.