https://www.zerohedge.com/news/2026-11-18/machines-started-attacking-each-other-and-your-bank-middle
the part about cyberattacks being involved in some of recent events, not the big ones, but lots of little ones that don't make it to the headlines.
every time someone patches holes, and announces it publicly, they notify attackers that anyone who hasn't already patched, can be attacked. the attacks are launched within minutes and the slow updaters get hit.
this seems to be a thing that is going to change the way software is delivered. there is two ways out of this, that i can see:
1. you take advantage of the fact that you can build big systems with few people using LLMs, and you aren't sharing your supply chain with anyone else. you eliminate upstream vulnerabilities by moving all possible code into your codebase, your codebase is private, and the network side at most reveals you are running code from AGPL'd repos, not that you are using them vanilla, necessarily.
2. anyone developing software that is deployed by many clients, must be the first to hear about the patches, and the public side doesn't appear until all of the users are already moved to the new codebase with the holes patched.
doesn't really seem to me like there's two ways about this. option 1 is the most decentralised. option 2 is such an amount of overhead and trust boundaries i don't see it being so popular anymore because the externalities now are the customers.