# Things about ARPA that can be improved DARPA Analyses can end up looking like hagiographies - this document is no exception. Because I’m trying to extract lessons, I don’t want to caveat every point. The productive thing to do is to call out the problems. * [[DARPA’s primary purpose is to develop military technology]] * [[The ARPA Model is high-variance]] * [[DARPA falls into many classic government organization traps]] * [[The ARPA model can lead to inbreeding]] * [[DARPA has a mixed record on transitioning]] * Contractors still have downsides * For fiddly modern systems that require UX they’re not good * DARPA is probably not super efficient * This is the other side of [[DARPA funds wacky things that go nowhere]] * DARPA often works on trendy things ### Needs more work * [[The ARPA Model is an unbuffered system]] <!-- #list --> [Web URL for this note](http://notes.benjaminreinhardt.com/Things+about+ARPA+that+can+be+improved) [Comment on this note](http://via.hypothes.is/http://notes.benjaminreinhardt.com/Things+about+ARPA+that+can+be+improved)