zygo_histo_morpheus @ zygo_histo_morpheus @programming.dev

Posts

1

Comments

261

Joined

3 yr. ago

Cleaning up your history before making a PR can make it easier to review so I always try to do that, but it's not something that should be over done because it can quickly eat up more time than its worth if you go overboard

The benefit with rebase is if you want to have maybe 4 commits insteadd of 10. When reviewing a large pr, I find that it's helpfull if it's broken up into a couple of coherent commits so that I can review it commit by commit. It's easier to follow the logic of why something is being changed if it's associated with a specific commit.

Sometimes squashing the entire commit is the right choice, in which case you can do what you're doing or use some built in feature to do that.

even a small amount of change into an LLM it turns out to radically alter the output it returns for huge amounts of seemingly unrelated topics.

Do you mean that small changes radically change the phrasing of answers, but that it has largely the same "knowledge" of the world? Or do you mean that small changes also radically alter what a llm thinks is true or not? If you think the former is true, then these models should still be the same in regards to what they think is true or not, and if you don't then you think that llms perception of the world is basically arbitrary and in that case we shouldn't trust them to tell us what's true at all.

Well if we have a reliable oracle available for a type of questions (i.e. Wolfram Alpha) why use an llm at all instead of just asking the oracle directly

The problem isn't just that llms can't say "I don't know", it's also that they don't know if they know something or not. Confidence intervals can help prevent some low-hanging fruit hallucinations but you can't eliminate hallucinations entirely since they will also hallucinate about how correct they are about a given topic.

Sometimes I'll solve a computer problem for someone in an area that I know nothing about by just googling it. After telling them that all I had to do was google the problem and follow the instructions they'll respond by saying that they wouldn't know what to google.

Just being experienced at searching the web and having the basic vocabulary to express your problems can get you far in many situations, and a fair bit of people don't have that.

Download a popular movie and keep your computer on for a while 🤷‍♂️

Although, seeding stuff that isn't popular is also important. I don't know what you're seeding but if no one is leeching maybe there aren't a whole lot of other people seeding either. When someone does leech, they might be very happy that you're there keeping that one torrent alive.

I can recommend fd to everyone frustrated with find, it has a much more intuitive interface imo, and it's also significantly faster.

Pair coding with vim is a skill in itself (for the vim user). You can make things a bit easier to follow by making liberal use of visual mode for example. I have a CoworkerMode command that turns on smooth scrolling via vim-smoothie and cursorline, and I've also added some stuff to the neovim right-click menu so that I can explicitly right click go to definition for example. It can be worth switching editor sometimes, but it's not always worth it if you're in the middle of something.

You can have qBittorrent running in mixed mode, which doesn't give you the privacy of i2p but does give you even more leachers than just using normal ip and helps grow the i2p network. Everyone should get i2p and use mixed mode or i2p only imo.

Dark UX is more like features that are intentionally misleading, enchitification is making your product worse in order to be able to make money of it.

UltiSnips is fantastic

Was it the version controlling or writing in markdown part that was a hassle?

One problem that's particular to node is that you can't unpublish packages if another package depends on them. As it says in the article, that means that no one can unpublish their packages, including the everyone package since someone apparently depends on that.

For work gitlab is fine, I'm sure your company can get the accounts verified for example. At least it's not microsoft

Bluesky has the most twitter like user base of all the twitter clones that I've tried, and it's up to you if that's a good or bad thing. It's not all segments of twitter though, there isn't really any of right wing twitter or crypto twitter for example (a lot of furries on the other hand) which is quite nice actually. It isn't really active or important enough to get a lot of the big drama or main character moments and there aren't really any celebs, journalists and politicians posting there. So it's a bit like twitter without many of the lows but also many of the highs.

Dorsey got bullied off bluesky by its userbase so there's that at least

I don't think I would use this actually, because I don't see how an AI could capture the performance. I'm a sub over dub guy anyway, but at least someone making a dub has a sporting chance to make an interesting performance.

Peak dishwasher is a great concept and I think it highlights something important in the way we think of technology. There's often this underlying assumption of technological progress, but if we look at a particular area (e.g. dishwashers) we can see that after a burst of initial innovation the progress has basically halted. Many things are like this and I would in fact wager that a large portion of technologies that we use haven't actually meaningfully developed since the 80s. Computers are obviously a massive exception to this - and there are several more - but I think that we tend to overstate the inevitability of technological progress. One day we might even exhaust the well of smaller and faster computers each year and I wonder how we will continue to view technological progress after that.