I saw a tool launch that does the job of one or two people in 5 minutes that would normally take them a full day. And we started with simple ugly image generators. That is incredible change and progress. It would be dumb to assume it stops there.
Discussion
It's also dumb to assume progress remains exponential, there are roadblocks that become very hard to overcome, by that logic we should have had a colony on Mars by the 1980s.
Not saying AI tech won't get better but we're seeing some real saturation with LLMs specifically.
For more info see "Scaling Laws for Natural Models" by Kaplan