It is not a lack of talent, it is a lack of good governance/administration that causes most software/IT failures for the government. USDS is actually pretty cool, and they seem to be doing really great work. But they're a small part of the system. In order to actually improve software/IT within the government we need to fix a lot of small problems.

Here are some more:

Just the universal waste within the way government treats software acquisition. If you ran a private company and you had small teams (each less than 10) going out and individually buying a software system, say half of them end up buying small-team licenses for Visual Studio as an example, you'd want to murder someone. They could save 20-40% by combining all purchases under a common contract.

Or instead of hiring an actual system administrator, the new guy gets tasked with managing in-house servers. Oh, but it's not one person for an office of 1k, it's every small team of 5-15 people dedicating one person to manage 1-5 servers. Servers that could, mostly, be shared across teams and so instead of 100 admins you could probably get by with 10-20 full-time and trained admins, along with a fraction of the servers and a fraction of the software cost.

The US government does this all the time because there is little-to-no communication involved. What's hilarious(ly sad) is that I've seen this where the purchases went through the same purchasing officer. A bunch of teams all buying expensive (because of scale) licenses when they could've combined them into one joint contract. Millions wasted a year doing crap like this even in smaller branches.

Fix the crap like that and then USDS's model can become effective and universal. Until then, there is so much waste abounding within the way US government does software/IT that it will always be a joke.