One of the biggest problems I've run into AI tooling is they brute force a solution instead of subconsciously thinking about things that engineers subconsciousl

One of the biggest problems I've run into AI tooling is they brute force a solution instead of subconsciously thinking about things that engineers subconsciously think about while crafting a solution, so even if it does manage to solve the problem you asked it solve it almost always isn't the best solution out there. The reality of problems such as this one is you don't discover them until the product is in a user's hand for awhile, so while LLMs may be useful is speeding up the resolution of these problems they're kind of useless at discovering them. Yet despite this CEOs of companies with major AI investments are promising an idealized future within a couple of years as though it hasn't already been nearly 3 years of ChatGPT. Despite all the hype and promises the pace of the product development lifecycle seems to remain unchanged. I think realistically it'll be at least 10 years before LLM-based tools and agents can do what CEOs are promising.