Linear and Salesforce both found validation, not code generation, bottlenecked AI workflows

Two engineering teams, Linear and Salesforce, independently published findings showing that AI-assisted coding created downstream bottlenecks in validation rather than in code generation itself. Linear reported in September 2026 that its CI pipeline struggled to keep pace after test suites nearly quadrupled, prompting infrastructure changes that cut pipeline time by around 34% and reduced typecheck time by 73%. Salesforce engineers Shan Appajodu and Ravi Boyapati documented a separate but related problem, where a roughly 30% rise in code volume led to larger pull requests and review latency that climbed quarter over quarter. The most alarming signal Salesforce identified was that review time on the largest pull requests plateaued, suggesting reviewers had stopped meaningfully engaging with oversized, context-heavy diffs. Together, the two reports point to a shared structural gap: as AI lowers the cost of producing code, the systems designed to verify that code have not scaled at the same rate.