Imagine every Rust program just calling unwrap on this result, and therefore the Rust library deciding to just never return an error. That’s analogous to what happened here.

LTS distros can track the LTS kernels and they’ve all been updated

What about return oriented programming?

Seems to me that “we ran the Jepsen test” is just something an LLM thought would sound good on a DB’s website.

It’s wild how bad it is. Not surprising for a vibecoded database, but why did they hire Kyle for a test, instead of first doing ANY testing themselves?

My best guess is their board/investors forced them to.

In this theory, without evidence just like their db claims, they were looking for external evidence to shut it all down or force replacement of key personnel.. And Aphyr understood the assignment.

I'd assume that no-one who works there really knows anything about databases; they may genuinely have thought that it worked. A lot of super-AI-pilled people seem to be deep into this "expertise doesn't matter any more" belief system.

No doubt because the LLM said it was ready to pass Jepsen.

^^ this is funny, but I think that's likely how it has happened.

And no doubt they wanted a headline to help boost the Anthropic IPO (I'm kidding but knowing the crypto crowd and the overlap with the AI crowd, who knows)

it is surprising even for vibe coded database.

Imagine bringing in any outside testing before your WHERE implementation works.

You can cut out the middle man and use https://github.com/marxarelli/masse, a CUE frontend for BuildKit.

However I don’t think anything using BuildKit is as fine grained as Bazel.

Dagger replaced BuildKit (a lot of reasons for this). The BuiltKit API sucks compared to Dagger's, so I personally prefer the middleman, but that is no longer the case. See:

https://github.com/dagger/dagger/pull/11856

Oh, TIL!

truth be told so did I, I knew it was coming but didn't know it had arrived, been deep in the agent / harness realm the last year

“Their tests showed that rqlite maintains linearizability of record insertion and modification, provided strong read consistency is enabled.”

The rqlite docs however tell you to never use Strong mode and that Linearizable is sufficient. https://rqlite.io/docs/api/read-consistency/#strong

This smells a lot like the situation MongoDB had in its early years that made it lose a lot of trust in the community.

Looking further into this: The Linearizable mode is newer than the test, so we don’t know if it would have passed.

https://philipotoole.com/faster-reads-same-guarantees-linear...

How are agents helping you? What was the trouble with Postgres before?

My guess is that coding agents expose deep documentation and advanced patterns better or least faster that wading through everything.

Unfamiliarity/unwillingness to deal with the infrastructure admin

Many CVEs have too high severity for what they describe.