>one line item would have been building compilers, which is difficult to even maliciously twist into working on weapons How so? Your compiler produces a binary

>one line item would have been building compilers, which is difficult to even maliciously twist into working on weapons How so? Your compiler produces a binary, binary gets loaded on a chip inside a weapon system. Is that not working on weapons? That wasn't so difficult. I get the overall point, military spending as a huge funnel of free money to be thrown at research, of which not necessarily all would end up in actual combat ready systems. But you don't know which of your research would the uniforms throw away in government labs and which they would put on assembly lines and send to the field. Compilers are extraordinarily essential pieces of software that lie downstream of literally any source program, so even if your compiler was a research toy, it can be scavenged for that one algorithm or tricky data structure that can increase the efficiency of production compilers that produces the binaries that animate the killing machines on the field. For example, here's [1] an article about how formal verification helps unmanned helicopters be more resistant to cyber attacks. I didn't even search for it specifically for this comment, I had it on open tabs for 3 days. [1] https://m-cacm.acm.org/magazines/2018/10/231372-formally-verified-software-in-the-real-world/fulltext https://m-cacm.acm.org/magazines/2018/10/231372-formally-ver...