I am staunchly anti-hype and a firm "whatevers" on the rest of AI. Like you said, it has its place. The biggest issue I have is that it is being pushed as another dopamine fix. I ain't gonna lie, turning an idea into code in 5 mins is really fucking cool. Unfortunately, its really fucking addictive so I give myself multi-day breaks after a day or two of coding and fucking around with my homelab.

(Tinfoil hat time) If anyone hasn't noticed, there are some clear distinctions between enterprise LLM tools and regular consumer LLM tools. The consumer-grade plans are more prone to random mistakes and forgetfulness. My theory is that more "mistakes" and resulting fixes not only burn more tokens, but also push dopamine levels higher and lower. Aside from coding blatantly obvious security issues, enterprise LLMs are much more reliable.

LLMs are a tools. You solve problems. If an LLM is solving all of your problems, you are the tool.