We never can have total trust in LLM output, but we can certainly sanitize it and limit it's destructive range. Just like we sanitize user input and defend with pentests and hide secrets in dot files, we will eventually resolve to "best practices" and some "SOC-AI compliance" standard down the road.
It's just too useful to ignore, and trust is always built, brick by brick. Let's not forget humans are far from reliable anyway. Just like with driving cars, I imagine producing less buggy code (along predefined roads) will soon outpace humans. Then it is just blocking and tackling to improve complexity.
It's just too useful to ignore, and trust is always built, brick by brick. Let's not forget humans are far from reliable anyway. Just like with driving cars, I imagine producing less buggy code (along predefined roads) will soon outpace humans. Then it is just blocking and tackling to improve complexity.