My experience is kind of the opposite of what you describe (working in big tech). Like, I'm easily hitting 10x levels of output nowadays, and it's purely enabled by agentic coding. I don't really have an answer for why everyone's experience is so different - but we should be careful to not paint in broad strokes our personal experience with AI: "everyone knows AI is bad" - nope!
What I suspect is it _heavily_ depends on the quality of the existing codebase and how easy the language is to parse. Languages like C++ really hurts the agent's ability to do anything, unless you're using a very constrained version of it. Similarly, spaghetti codebases which do stupid stuff like asserting true / false in tests with poor error messages, and that kind of thing, also cause the agents to struggle.
Basically - the simpler your PL and codebase, the better the error and debugging messages, the easier it is to be productive with the AI agents.
What I suspect is it _heavily_ depends on the quality of the existing codebase and how easy the language is to parse. Languages like C++ really hurts the agent's ability to do anything, unless you're using a very constrained version of it. Similarly, spaghetti codebases which do stupid stuff like asserting true / false in tests with poor error messages, and that kind of thing, also cause the agents to struggle.
Basically - the simpler your PL and codebase, the better the error and debugging messages, the easier it is to be productive with the AI agents.