Questions for those who insist on reviewing all code: Why do you not the output of the compiler? Do those reasons not apply to agentic output?
I think a group of us have found, for certain domains, the agentic output is passing all of our standards we have for correctness (ad hoc tests, automated tests, etc) so we don't review the code, just like we don't review the assembly or bytecode. The compiler and assembler have bugs, and its more likely the agent system has a higher probability of bugs. But it's still good enough for personal projects, web apps, prototypes, internal tools, analytical helper tools, etc.
I think a group of us have found, for certain domains, the agentic output is passing all of our standards we have for correctness (ad hoc tests, automated tests, etc) so we don't review the code, just like we don't review the assembly or bytecode. The compiler and assembler have bugs, and its more likely the agent system has a higher probability of bugs. But it's still good enough for personal projects, web apps, prototypes, internal tools, analytical helper tools, etc.