veselin

The trick is that now one will depend on somebody who is not boost for these nice mappings. I brought such a dependency for gRPC once and now it doesn't work with bazel 1.0. I either have to find the author, understand all the bazel trickery, or switch to something else, because most of these bindings depend on many bazel features.

So, bringing such multiple third party bandaid is currently not such a great idea. It would be a bit better if boost itself provided it.

This is a not very difficult problem to solve. Whenever there is a dispute, send somebody to verify the property. For sure AirBnb already has an anti-abuse team, so it is an extra step there. Also, disputes anyway cost them money for the person that answers emails, sometimes do refunds, legal risk, etc. So it will also not be more expensive.

It is one of these "growth" problems that a company is good to have and normally solves in an easy way. Furthermore, the author is still an active user of AirBnb.

As far as I remember, when websites were doing window pop-ups and Google toolbar removed them, they didn't block the toolbar users. So, the risk is low... unless Firefox does something stupid like full ad-blocking.

But if something like this happens, I will be happy to build/use an extension for the other browsers that makes such websites invisible in all major social networks.

I am surprised that the text mentions the word contraction and yet doesn't really do https://en.m.wikipedia.org/wiki/Contraction_hierarchies

Using something like this will on average get pathfinding for something like O(LogN) time. May sound strange or impossible at first - this is less than the number of squares the path goes through, but works and does not need huge space overhead. This is because the CH algorithm selectively adds shortcuts to the graph essentially making it completely skip many squares. Also this is not a heuristic, but a provably correct algorithm (no provable time complexity for arbitrary graph, but I think logN or maybe log^2(N) can be established for the simple graph in this pathfinding).

We may keep arguing about what is better. But even if I accept your argument, we will conclude that there was a change for the sake of change. Not something that the users wanted.

UX: no, you have to take the bad and non-intuitive UI.

the three buttons no longer work as before - e.g. long press no longer does split-screen. Furthermore, the UX before was cards one on top of each other, the open app is the topmost. The current "logic" is with cards sideways. Which is ugly, non-intuitive and copied from iOS. If you move left to take another card, suddenly your previous card does not remain on the right side. How is this intuitive? Even from business perspective this does not make sense, because once I got used to the new non-intuitive and less efficient UI, the switching costs to iOS for me will be lower.

The article is written by implying the wish of the author is the wish of the users. No, users don't want updates. They get annoyed.

I am on Android 10 and honestly I wish I didn't update. The update bricked by phone until I went in the bootloader to reset it to delete my data. Was it great to have it on day 1 (essential ph-1)? Maybe 5 years ago I would say yes, but now no. It wasted 2 hours of my time. Next time I will buy a phone with a good camera instead.

My experience with Android 9 was similar, but didn't break anything. The UI changes were just not good. I personally know the reason (worked with some of them) - all (unfortunately really all, no idea why this wasn't fixed) designers at Google are Apple users and have a bit if Apple envy. Apple can change the UX for everyone and make users get used to it (even if it is less efficient for everyone, which it is). Why wouldn't Google be able to do it? So Android had the best app switcher, but converged to the iOSish variant. I wish Google would just ignore Apple like Microsoft does it in their Windows.

In a similar way, most journalists are Apple users and there is one "right" way to do things. So we get this nonsense in articles and blogs.

I question all these scores. The reason is that the Apple chip also has pretty low frequency and not so fast memory. This would mean that there is a lot of IPC opportunity left in comparison to the top x86 chips. This is the range of 2x in comparison to Skylake. This is a crazy wide core. What kinds of reorder buffers are required for this? Just 2x more execution units won't be enough.

I have "beaten" it. Play initially like you would - typically a human would change his choice often. The network would learn from you and gain a lead of say 10 points. After this, if you pick something that gives you a win start sticking to it while it gives you a win. You will quickly regain the lost points and gain a small lead. This seems to work every time I try it. After 60 moves I always lead a bit. Btw, the game gets slow then.

Not sure a bigram or trigram model would not do better than the neural network.

Most people in Bulgaria are more concerned that the breach gives new weapons to scammers. There are a handful of groups that scam old people to transfer money pretending to be their relatives in need. There are groups that steal properties or even companies by doing a transfer with fake documents (then the property is resold of course).

The finance minister said on the TV these are anyway illegal activities with or without the leaked data. The problem though is that nobody was imprisoned for such crimes for ages, just like there are no politicians going to prison in the last ten years. Guess how many people believe him now.

This one is a character level language model. The Microsoft one is per token.

This one should be the closest to the Microsoft model: https://miltos.allamanis.com/publicationfiles/allamanis2018l...

Btw, simple neural networks are far from doing the best for code. The kind of accuracy LSTM gives is already achieved with good n-gram models. E.g. see this one on a Linux code prediction: https://arxiv.org/pdf/1506.02078.pdf Also, there are other techniques that improve on top of the n-gram models: https://files.sri.inf.ethz.ch/website/papers/charmodel-iclr2...

This is indeed a very interesting topic. I am working in the area and we get some rule-based models perform as good as neural networks and overall there are good steps: http://www.srl.inf.ethz.ch/papers/charmodel-iclr2017.pdf

These are models that can be inspected by an expert (or even non-expert) and then analyzed. But I guess that we must go past the neural network hype. At the moment it is very common that if I mention machine learning to people, they correct me and say "deep learning" :)

One thing that is not clear to me is if OOP code such as copy constructors will benefit from using the vector registers. Just having vector copy of objects may turn out to be a big difference.

Otherwise examples with array of size 4 elements are nice to show, but in most of these cases it is also easy to use intrinsics.

We tried to evaluate on data that is as much as possible independent from the training data. So we evaluate on projects outside of github.

It never shows types of return values, unless it is undefined. The reason is that the method may be overridden with a method with other return type.

Yes, it is statistical. A browser extension would be cool!