@bagder@mastodon.social this is by far the most important lesson I have been trying to teach my engineers and product managers. I have had so much success using LLMs to do parts of my job, and so others try to emulate with far less luck. What they don't see is the process I've developed to only use the LLM for things I know it is good at, or the time I spend reviewing everything it produces to ensure that it is a level of quality that I am happy with. Definitely a time saver and lets me get more done, but not a tool that can run unsupervised.
I still can't fathom throwing it at an unfamiliar problem or codebase and blindly trusting the results, but unfortunately the AI marketing has convinced so many people that you can :raccoon_junk: