Anecdotally, for small tools and projects, it can crank out functional tools that would take me days or weeks to remember how to do manually. Every time I ask LLM to write something, it also writes tests and validates its code against known values. I additionally spot-check the LLM's code against my own known values in prod, and have had to correct it on some functions. The code I write is mainly analysis of logs and configurations, and they seem well trained on the products that I am using.

Testing. It's all about having unit tests, and known outcomes that the LLM can check against every time it makes changes. For my work, that is good enough. For safety-critical systems I would demand more rigor in the testing and the human review and creation of tests.