AI coding agents generate more code, but not more software

via Ars Technica · 15 views · 1w ago

AI coding agents are producing more code, but not more software. A new study reveals a fascinating paradox: while these agents are churning out lines faster than ever, overall software development isn't speeding up. The reason? Human review. Developers still need to meticulously check and integrate that AI-generated code. This human oversight creates a significant bottleneck, absorbing any efficiency gains. So, for now, AI is boosting output, but not necessarily productivity in software creation.

← Back to the feed