Weeding is among the oldest problems in agriculture and one of the least changed in principle: unwanted plants compete with the crop for light, water and nutrients, and they have to be dealt with. The industrial answer for most of the last century has been to treat the whole field, because treating the whole field is the only thing that could be done at scale. A person can distinguish a weed from a crop plant effortlessly but cannot cover a large field repeatedly; a sprayer can cover the field but cannot tell one plant from another. That trade-off - precision or scale, never both - has shaped how fields are managed far more than most people outside farming realise.
Robotic weeders change the terms of that trade-off, and this is the genuinely significant part rather than the novelty of a machine in a field. A machine moving slowly along the rows, looking at the ground continuously and deciding plant by plant, can act on individual plants across an entire field. What it does about the weed varies - some remove it mechanically, some apply a very small, targeted dose exactly where it is needed, some use heat or light - and the differences matter less than the shared property underneath: the unit of decision has changed from the field to the plant. Scale is no longer paid for with precision, because the machine's attention does not get tired and does not cost more per plant when there are more plants.
The consequences follow from that and they are practical rather than futuristic. Treating only the plants that need treating can use a fraction of the chemical that treating a whole field uses, which is a direct cost saving as well as an environmental one, and it reduces what ends up in soil and water. It offers an option to growers who cannot use chemical control at all, for whom the alternative is hand weeding at a cost that limits what can be grown. And because the machine is looking at every plant anyway, it accumulates a detailed record of what is actually in the field - which is information a grower has never been able to get at that resolution, and which may in time prove as valuable as the weeding.
The honest position on where this has reached is that it is real but narrow. These machines are working commercially in high-value row crops - vegetables and salads, where plants are in tidy rows, the crop is worth a great deal per hectare, and hand weeding is expensive enough to justify the equipment. They are not yet general-purpose, they are slow, they are costly, and they do not deal well with fields that do not look like the fields they were built for. The direction is nonetheless clear and the principle transfers well beyond farming: when the cost of paying attention to each item individually falls far enough, treating everything the same stops being the sensible default it has always been.