It seems highly plausible that you could counterfactually affect many more acres of this land (and thus many more soil animals) through building houses or other structures than trying to maintain factory farms.
This would not necessarily undermine your overall argument but, interestingly, Tomasik’s (2016-2022) estimates seem somewhat in tension with this claim. According to him, it’s really hard to beat “buying beef” in terms of cost-effectiveness to reduce wild invertebrate populations.[1] (Not saying I agree or that I think we should reduce wild invertebrate populations.)
Thanks, Jim. I forgot to comment on that. When I read the post, I immediately thought there is no way building houses increases paved area more cost-effectively than buying beef increases agricultural land. I estimate that buying beef increases agricultural land by 51.6 m²-year/$, which means 1.94 $ (= 100⁄51.6) of beef would increase it by 100 m²-year. So a building whose base was a square with 10 m (= 100^0.5) of side, and whose construction was accelerated by 10 years would have to cost 19.4 $ (= 1.94*10) for it to increase paved area as cost-effectively as buying beef increases agricultural land. Of course, such a building would be way more costly. Here is another check. It looks like the construction cost in Portugal is 2.7 k€/m², 3.11 k$/m² (= 2.7*10^3/0.868), and I think this accounts for the area across all floors. For 3 floors (0, 1, and 2), the cost based on the paved area would be 9.33 k$/m² (= 3.11*10^3*3), or 1.07*10^-4 m²/$ (= 1/(9.33*10^3)). Assuming the intervention accelerated the construction by 10 years, it would increase paved area by 0.00107 m²-year/$ (= 1.07*10^-4*10), 0.00207 % (= 0.00107/51.6) as much as I estimate buying beef increases agricultural land. I am somewhat surprised ChatGPT 5.1 got it so wrong in thinking that building houses could be a great way to change land use.
I appreciate you taking the time to respond to my post here and elsewhere. I am personally not surprised at all that GPT 5.1 got this type of thing wrong. Like I said in the post, I don’t believe LLMs should be trusted on complicated and important questions like these, and I only used GPT because Real Estate Development was a somewhat arbitrary example, not a specific call for focus.[1] My real CTA with this post is for people who believe nematodes (and other soil animals) are our primary concern to investigate other interventions that don’t concern factory farming. YGG’s comment captures it well.
There could be some interesting discussion about the difference between buying beef and spending money developing land in a way that could turn a profit allowing you to earn to give, meaning a straight cost-effectiveness comparison doesn’t really make sense.
However, I really don’t really know if Real Estate Development specifically is a better or worse use of your time. Again, it was an arbitrary example.
This would not necessarily undermine your overall argument but, interestingly, Tomasik’s (2016-2022) estimates seem somewhat in tension with this claim. According to him, it’s really hard to beat “buying beef” in terms of cost-effectiveness to reduce wild invertebrate populations.[1] (Not saying I agree or that I think we should reduce wild invertebrate populations.)
Although he omits the fact that agriculture might in fact increase soil nematode populations, as also pointed out by Vasco in another comment thread here.
Thanks, Jim. I forgot to comment on that. When I read the post, I immediately thought there is no way building houses increases paved area more cost-effectively than buying beef increases agricultural land. I estimate that buying beef increases agricultural land by 51.6 m²-year/$, which means 1.94 $ (= 100⁄51.6) of beef would increase it by 100 m²-year. So a building whose base was a square with 10 m (= 100^0.5) of side, and whose construction was accelerated by 10 years would have to cost 19.4 $ (= 1.94*10) for it to increase paved area as cost-effectively as buying beef increases agricultural land. Of course, such a building would be way more costly. Here is another check. It looks like the construction cost in Portugal is 2.7 k€/m², 3.11 k$/m² (= 2.7*10^3/0.868), and I think this accounts for the area across all floors. For 3 floors (0, 1, and 2), the cost based on the paved area would be 9.33 k$/m² (= 3.11*10^3*3), or 1.07*10^-4 m²/$ (= 1/(9.33*10^3)). Assuming the intervention accelerated the construction by 10 years, it would increase paved area by 0.00107 m²-year/$ (= 1.07*10^-4*10), 0.00207 % (= 0.00107/51.6) as much as I estimate buying beef increases agricultural land. I am somewhat surprised ChatGPT 5.1 got it so wrong in thinking that building houses could be a great way to change land use.
Hi Vasco,
I appreciate you taking the time to respond to my post here and elsewhere. I am personally not surprised at all that GPT 5.1 got this type of thing wrong. Like I said in the post, I don’t believe LLMs should be trusted on complicated and important questions like these, and I only used GPT because Real Estate Development was a somewhat arbitrary example, not a specific call for focus.[1] My real CTA with this post is for people who believe nematodes (and other soil animals) are our primary concern to investigate other interventions that don’t concern factory farming. YGG’s comment captures it well.
There could be some interesting discussion about the difference between buying beef and spending money developing land in a way that could turn a profit allowing you to earn to give, meaning a straight cost-effectiveness comparison doesn’t really make sense.
However, I really don’t really know if Real Estate Development specifically is a better or worse use of your time. Again, it was an arbitrary example.