CMV: LLMs are fantastic if the person using them is competent.
State of the art LLMs can save a lot of time researching, writing and coding if the person using them has basic knowledge of the field and critical thinking skills.
All SoTA LLMs extensively cite their sources if prompted to -making it easy to validate the arguments they are making- and reading through and checking generated code + writing is much faster that making it from scratch.
So many people complaining about LLMs being useless or causing technical debt only get bad results because they are too lazy to read through and check the work it produces or don't have basic research or coding skills necessary to catch when they make a wrong turn or be specific enough about what they want in their prompts.
TLDR: If you think SoTA LLMs are bad it's a skill issue.