AI Responses Are Often Verbose, Contain Plausible-sounding Errors, and Are Packed With Jargon, Making Them Harder to Read Than a Direct

The critique highlights that AI responses are often verbose, contain plausible-sounding errors, and are packed with jargon, making them harder to read than a direct human reply. The author recounts receiving a Claude-generated sentence so dense that nearly every word required a dictionary lookup. The advice: use AI to draft, but always read, understand, and validate the output before rewriting it in your own words.

That effort, the author says, is where real value lies. The post also takes aim at code review, noting that developers can now ship code by pasting a ticket description into Claude Code, ignoring the code itself, and then feeding reviewer feedback back into the tool. In such cases, the author argues, the reviewers are effectively doing the implementation, with the original developer acting as a mere conduit.