“Security theatre” is what I’ve named the contact in my work phone for the call center I have to call every time I accidentally use the “one time password” more than once (because god forbid they implement proper SSO, meaning I have to do a shotgun login run every morning). When I call them all I tell them is my name and that my account is locked.They click a button and we’re back. Complete waste of time on everyone’s part.

Abstractions aren’t concrete and all of these standards you’re referring to are concrete data serialisations. You may be interested in CUE which captures this concept in its design.

You’re doing it right by avoiding as much of Gitlab’s CI features. I’ve seen versions where scripts are inlined in the YAML with expressions in random rule fields and pipeline variables thrown all over the place. And don’t get me started on their “includes” keyword, it’s awful in practice, gives me nightmares.

Then I write a Kubernetes manifest in YAML with JSON schema validation and the heart rate goes down again.

I agree. You can’t just dismiss the problem saying it’s “just data represented in vector space” and on the other hand not be able properly censor the models and require AI safety research. If you don’t know exactly what’s going on inside, you also can’t claim that copyright is not being violated.