Like... how easy is it to not know how anything works and generate a "working" program, using the loosest possible definition of "working", using LLMS?
Generally I find AI autocompletion in cursor pretty bad with Odin, and Claude generally gives me code with incorrect syntax. I think the language is too new to be effectively used with AI assistance.
Now imagine how many normie, computer-illiterate federal employees in fairly sensitive roles have had various credentials leaked over the past few years.
There are safe guards for information not to leak. Those safe guards make it very hard to get the info, not impossible, but very hard. Walking into a government office and plugging in your personal Macbook, and running whatever software you want with "god" powers on the network makes it a lot easier to gain access to whatever data is required. Even if its unintentional (big if) from the DOGE's side, at this level you are target by state actors and they will get to your personal devices if they want.
I worked in Federal government on classified systems. There were many safeguards in place, most importantly networks that were 100% disconnected from the Internet and locked down workstations. That made sure that even the most inept user could not cause a problem like this.
Everyone I worked with respected OpSec and would never do something as risky as bring in an outside laptop and connect it to the network. DOGE has been so reckless that I believe they wanted to have the system hacked, because seeing our government destroyed is their real objective.
