Cognition’s Scott Wu says AI coding agents shouldn’t replace humans

Wu's statement arrives after Devin faced significant public criticism over whether its real-world performance matched its launch-era claims, making this less a philosophical stance and more a response to a credibility problem. The augmentation framing is doing defensive work here, not just positioning work.

Modelwire has no prior coverage to anchor this to directly, so this sits largely disconnected from recent activity in our archive. More broadly, it belongs to a recurring pattern across the coding-agent space, where vendors (Cognition, GitHub Copilot, Cursor, and others) have cycled between capability maximalism and reassuring 'humans stay in the loop' messaging depending on the public mood. The displacement anxiety Wu is addressing is real and has drawn attention from labor researchers and policymakers, but a founder interview is a thin vehicle for resolving it. Without a product update, a methodology change, or third-party validation attached to this statement, it is difficult to separate genuine strategic recalibration from damage control.

Watch whether Cognition publishes updated, independently reproducible benchmarks for Devin within the next two quarters. If the augmentation framing is backed by concrete workflow data showing human-plus-Devin output versus solo developer output, the repositioning has substance; if it remains at the interview-quote level, it is messaging without a product story behind it.