Cursor Agents Raise Task-Supply Questions

The big number is mostly an orchestration story, not a magical infinite backlog. Teams get scale by turning one real change into many small jobs and by routing every review, test, refactor, and cleanup step through agents.

• –Work usually comes from existing sources: bug reports, support tickets, failing CI, feature specs, code review feedback, and roadmap items
• –The trick is decomposition: one feature becomes a spec pass, implementation, test generation, doc updates, lint fixes, and verification
• –Agent volume rises fast when companies run parallel pipelines across many repos, squads, or branches, even if each individual task is narrow
• –“Thousands of agents” is often aggregate throughput, not thousands making autonomous product decisions at once
• –The real bottleneck is quality control: without strong evals and human checkpoints, more agents just scale churn and hallucinations
• –The Reddit discussion also reflects a common skepticism: some of these claims are likely marketing shorthand for a much more constrained workflow