pammf

It all depends on the use case…I have two raspberry units running as exit nodes back in my home country, one in my mother’s place and another in my in-laws’. They have regular internet providers routers, and at least one of the routers wouldn’t even be able to properly support port forwarding.

Tailscale allowed me to setup everything at home and just plug them to their network in 5 mins.

Even with the same model results are different. I work in a startup where they went bonks with AI ~1 year ago. Every employee got a Claude account that I don’t even know what tier is (we just have what feels like unlimited tokens), they implemented infamous AI usage leaderboards, people were getting “you’re not using AI enough” feedback and all of that…

Over time, it was impressive to see how much the output of the same prompt would differ between us, and it seems to be heavily influenced by the role of the user. I.e. answers coming from the agent of a Data Engineer, a Backend Engineer and a Domain Researcher would not be the same at all. Some of us use harnessing tools, some not, but regardless it we saw differences.

So, for example, when a backend engineer would send a question to a data engineer, there was *some* added value even when it was a lazy copy & paste.

Agree that it isn’t nice to just copy & paste, but the assumption that your AI generated answers will the same of someone’s else is wrong. Everyone using AI for a while built a different set of preferences, memories and context in general which influence the answers.

AI is here to stay. Like it or not.

We will still see several reports of over adoption, mistakes, regression… all will only serve to learn, refine, and hopefully regulate.

I think it’s pretty naive to expect the entire world will simply discard the technology and go back to having humans doing it all.

Trump will go down in history as the clearest marker of the United States’ decline.

In Claude Code, it always shows the diff between current and proposed changes and I have to explicitly allow it to actually modify the code. Doesn’t that “fix” the copy-&-paste issue?

Real reason is humans are way too optimistic in planning and, for some reason, tend to overlook even more rare, but catastrophic risks.

I’m almost sure that the system had some sort of local replication and versioning that was enough to deal with occasional deletions, rollbacks, and single non-widespread hardware failures, so only the very catastrophic scenario of losing all servers at the same time (that for sure wouldn’t happen anytime soon) was uncovered.

> If, as a manager, you want a more structured means of conveying institutional knowledge generated within such channels then it's on you to put that together

Isn’t the GP describing exactly that?

Such as?

Would it switch the lights if you’re only listening with headphones while your camera is open?

Most of my awkward moments happened with family members casually cruising on my background while someone else had the word. Usually they can hear when I’m speaking and don’t enter the room.

You are right. My comment wasn’t meant to completely invalidate the point of the article or to provide an alternative exhaustive list of causes, but more to bring this other aspect that I felt wasn’t surfaced yet.

A good part of that is a disguised sense of superiority.

Life becomes way lighter when you realize other people are also smart and what you’re “fixing” can very likely be:

- something so unimportant that no one felt it was worthy working on

- something that was supposed to work like that, and you simply don’t agree and want to make it your way

Well, from outside it seems that the list of friendly nations keep getting smaller and smaller… and good luck replacing the vast majority of your supply chain with local workers. Just look where all the products you currently have were made.

Iceberg has the hdfs catalog, which also relies only on dirs and files.

That said, a catalog (which Delta also can have) helps a lot to keep things tidy. For example, I can write a dataset with Spark, transform it with dbt and a query engine (such as Trino) and consume the resulting dataset with any client that supports Iceberg. If I use a catalog, all happens without having to register the dataset location in each of these components.

I share the sentiment, and I also get downvoted every time I post something on those lines here. Only thing that can make you even less popular is saying that using cloud services can make sense depending on your priorities. =D