"Reliably Mechanical" is a phrase that resonates a lot with me in regards to tech.

I think the success of technology has always rested on two axes. One is speed. Doing things faster is always appreciated, because we can do more. The other, perhaps more critical axis is 'correctness'. Good technology doesn't just do things faster than people, it does them more reliably, more precisely, and accurately. The cost of mistakes tend to grow exponentially the further they are in the chain of production, as a mistake in the next step ruins all the productivity of the previous step. So, speed is usually secondary. The history of "The first patented machine to do X" is littered with failed designs that were too inaccurate, too unreliable. The first design do combine speed with correctness is the real revolution.

This is how I feel about the LLM stuff; it's fast but not reliably correct, or "reliably mechanical". I think the Genie is out of the bottle so to speak and we're stuck with this stuff, but once the venture capital gets tired of pitch-forking piles of money into a furnace, I think you'll see mostly local models and the big ones will be too expensive for the average person beyond whatever google or other is willing to lost-lead for sweet ad revenue or to create propaganda realities (ala Grok).