I was in the same situation 6 months ago. Since then I’ve fully adopted an AI driven workflow and it would be hard for me to go back. It’s not perfect, but it’s a great assistant. It does make slightly concerned about my job prospects in the long term so I hope they hit a plateau in the long term

Or maybe it’s all those US based btc ETFs.

This will be Bluesky’s trial of fire. I’ve noticed a significant increase in bot accounts lately.

How large is the disk footprint at the moment?

I want a feature to block all people without a private domain handle. That will install cut spam.

I see. Thanks for the clarification. So why wouldn’t you be able to replicate just a subset of the network then, if each user has their own merkle-tree? Perhaps there are parts of the state which are global (like followers), but you should be able to index only what you’re interested in.

Haven’t been getting the email codes, and I tried several times with different emails. I am running my own pds, could that be an issue?

Look into merkle search trees and prolly trees. Nice primer here joelgustafson.com/posts/2023-0...

Yeah, so you’d have to pull that content too. It’s still a subset of the network. I thought that merkle search trees, which are the foundation of atproto, enables this partial synchronization.

Why can’t relays relay only a subset of the network based on the interests of their users?

Be careful, it might turn into Nostr 😀

I’m a fan of atproto and blueksy, but that “temporarily” is pulling a lot of weight. It all depends on how temporarily this will stay centralized. Hopefully the app views and relays get more decentralized.

Fully agree. Many teams would be much better served by ecs fargate or google cloud run.

Classic flawed argument against Kubernetes.

Yes, so when something goes wrong I can dump the whole relationship tree of resources and see what is not in ready state.

Yes, I know about owner references but unfortunately its use is quite limited. For example, when a pod mounts a volume from a secret created by a certificate, I would like that relationship to be inferred automatically. ArgoCD keeps its own tracking. I would like this to be added to k8s directly.

Resource relationship tracking, but not through optional annotations like some projects do now, but as a core feature of k8s. I would like to be able to have dependency trees that aid in debugging issues.

Do you run your own Bluesky pds?