the short of it is just that if there is no linear increase in yeild, your tech product will probably fail. Most of this nonsense comes from tech bros who have never farmed and don't have the common sense to ask a farmer what they actually need.
Incorrect. I worked for an Ag tech company for almost a decade. Ten years ago, farmers were downloading high resolution satellite imagery of their fields directly to their GPS enabled precision sprayers so that the spray rate would adjust continuously based on the vegetative index of the land they were over. I don't know what state of the art is today, but I imagine it would surprise folks who aren't in the field.
Vision systems for weeding are effective. Instead of spraying everything, cameras look down as the sprayer is pulled behind a tractor, and only weeds get sprayed, or zapped, or hammered. Uses far less pesticide. Deere and others sell this. Deere prefers to sell this as a service, where farmers pay fees when the vision system is enabled.[1] There are a bunch of autonomous robot startups with similar systems, but none seem to have grown much.
Tomato picking robots have been demoed since at least 2018, but still aren't used much. There are at least half a dozen startups. It's not hard to do with off the shelf robots, but not cost-effective yet.
Automated cow milking is used in New Zealand and Australia. Those countries have extensive agriculture but not much of an underclass, so they have to pay workers real money.[2]
Automated meat cutting is also used in those countries.[3] Mostly lamb, which isn't a big thing in the US. Fully automated beef lines don't seem to be available yet, although the company that builds the lamb systems is getting there.
Vision-based sorting is automated, fast, and cheap.[4] That's why when you buy packed berries, there are no bad ones any more.
All the big field crops - corn, wheat, hay, soybeans, cotton - were mechanized decades ago, of course.
What else is actually working?
[1] https://www.deere.com/en/technology-products/precision-ag-te...
[2] https://www.youtube.com/watch?v=o515XdtU7NM
Ag is much more than farming, and farming is much more than dirt.
Ag is very much ready - past ready - for disruption. The biggest blocker there is that "traditional" SV startups can't make a dent; their business models don't suit it well. Farmers - even mega-corporate ones - are cautious. It's a mature market with low margins, so they aren't going to make a capital investment in what they see as unproven technology.
To succeed in agtech you pretty much have to exude the opposite vibe that startup founders usually have.
Early 80's, in college, we had the Cyber mainframe, and PDP 11/70. The Cyber was used for computer classes (comps sci [Fortran, Pascal], infosystems [COBOL]), engineering classes (SPICE, Aeronautical Engr had some Fortran IV stuff the had to slog through), statistics runs (SPSS - lot of the sciences used that)), the school had some back office software running on it (you noticed during registration, the machine was noticeably loaded). PDP was used mostly for introduction CS classes, and it had a public game account (very nice Star Trek program on that one).
Outside of a TERAK and a Textronix 4050 series computer in the Math lab, there were no real microcomputers on campus, no "public" ones used for teaching.
The first micro computer lab on campus, was in the Agriculture school. They had a full lab with 9 Apple ][ computers, all running, I guess, a sophisticated software package for computing feed and what not. We were there to play Wizardry, of course, my good friend was the lab tech.
But it was just interesting how the Ag school was a pioneer in the personal computing space at an, ostensibly, school known for its science and engineering programs (though it was also know for its Ag and Architecture schools).