Let me introduce you to my baby, Belle.

I recommend wrapping the git cli commands using subprocess, using porcelain output modes etc, and parsing the output.

We have had stability problems with GitPython (which wraps gitdb). On Linux gitdb does clever things with sliding mmap, which caused some crashes (in a multi threaded environment), and I found simple race conditions in the code for writing loose objects, which is about as simple an operation as can be, so I lost faith with it. I do use gitdb in one read-only single-threaded system; it’s undoubtedly fast.

The biggest issues with git libraries are around the complexity of git configurations. Any independent reimplementation is probably going to support the most common 99% of features but that 1% always comes back to bite you! We use a lot of git features in service of a gigantic monorepo, like alternates and partial clones and config tricks.

If we use command-line git we get 100% compatibility with all git configuration and ODB features, and it’s hard to ensure that with an independent git implementation (even libgit2).

When you say “that solution doesn’t scale well” - we have made it scale. git itself scales well for operations it can perform natively, you just have to use the features effectively, often the high-level operations but sometimes lower-level commands like git cat-file --batch, git mktree --batch, etc. It’s not as fast as gitdb but fast enough, and I can have high confidence that I can write something once and it won’t break or cause problems later.

I doubt they are using Johansson’s voice. I expect they need much more studio-quality training data than they would have for her.

The desire to create a “Her” might be real but explains why they chose a similar voice actress, made Sky the default, and continued to pursue Johansson to some day create the real thing.

Suspending the Sky voice looks guilty but it might be a temporary action while the legal team considers their response. There might be a non-zero risk of being found liable if there were directions in the voice casting process to seek a result comparable to Scarlet Johansson. You’d want to collect and assess correspondence to see if that’s a possibility, which might take a while.

Yeah, decorate it just with a tremendous amount of dark red paint, spattered away from the fan, heaviest in the fan corner

I don’t deny that this kind of thing is useful for understanding the capabilities and limitations of LLMs but I don’t agree that “the best match of a next phrase given his question, and not because it can actually consider the situation.” is an accurate description of an LLM’s capabilities.

While they are dumb and unworldly they can consider the situation: they evaluate a learned model of concepts in the world to decide if the first word of the correct answer is more likely to be yes or no. They can solve unseen problems that require this kind of cognition.

But they are only book-learned and so they are kind of stupid about common sense things like frying pans and ovens.

I think it’s fair to colour seasons and episodes with different scales because they are measuring different things.

Due to the Central Limit Theorem, average of 20+ episodes will have a smaller standard deviation than individual episodes.

For example, an individual episode with a score of 6 you’ll probably watch. A whole season with a score of 6, maybe not.

Sweet lemonade, mmm, sweet lemonade…

I’ve used it, but only when I’m on some really sketchy unencrypted WiFi network, like in an airport or a hotel.

It doesn’t offer location spoofing like all other VPNs on the market, which I would have more use for.

Yeah but it sounds less cool if you say thirteen point three millilitres than to say Four (4) TRIOS™!!!

Unless that person happens to be with their dad, that would then require finding their dad. That’s a whole extra person to find. It might be easier to skip straight to finding their dad.

Wait, what do Americans use? Only Signal and Telegram?

Use after free, null pointer dereference, double free.

Solutions to these in C end up looking a lot like Rust.

Rust has a lot going for it beyond just the safety thing: excellent package manager, powerful trait system and generics, helpful compiler errors.

The whole language is designed to help you avoid making the programming mistakes people tend to make, not just the borrow checker and memory safety.

Yes, Windows perfectly serves the purpose of running an SSH client to log into a Linux box. Totally adequate experience.

Finally, the drivers will be able to play Angry Birds while they race.

Yeah, I shell out for the premium electricity, the 99% electrons. The 95% stuff is fine but I have a lot of expensive devices; I want them to run as fast as possible.

Isn’t the point of these discounts to allow the store to unload goods that customers otherwise would avoid due to low remaining shelf life, and recoup some of the purchase costs? If they don’t discount it will end up costing them money.

Python already has this.

Maybe I’d take the first year in prison for a half a million* but ask me again 1 year later, do you want to do a second year? No, I want 600k. Next year, 700k. 44 years? Honey, you can’t afford it.

* I wouldn’t, my number is higher than that.

The price difference is quickly made up for with the re-usability factor.

I don’t think that’s true, CD-Rs cost pennies. You have to rewrite every CD-RW 4 or 5 times before it’s comparable in price. In practice, across every CD-RW ever made, the approximate number of times it is written is probably about 0.5