I have heard way too many “performance in the ballpark of C”, most of the time it means in some cherry-picked example it is slightly slower than C, and most program is 20-50 times slower.

The language design honestly remind me of the old PHP: uses hashtable and array as primitive data structures, and free memory at the end of a function to achieve memory safety.

It seems quite unbelievable to me that it is gonna have C-like performance, since hashtable is usually quite slow compared to even heap access (direct stack allocation, of course, is the fastest); but I would be happy to be proven wrong.

Not disagreeing with you, but I find it funny that this is the only data structure I have not heard of in this entire thread 🤣

I don’t know if Haskell etc affords you such control

You can have immutable arrary with vectors, but to mutate them you will need to wrap your action in a Monad. It even supports unboxed values.

https://hackage.haskell.org/package/vector

But I agree boxed default actually causes a lot of performance overhead in many high-level languages.

I heard somewhere people just elected a billionaire sex offender (in 2024), I imagine life must be swell over there.

You can also put tomato and salad in a pan, but there is no tomato and salad, and that is sad.

I feel yields different result than 5 ∪ 7 in the classical set theoretical encoding… I believe 5 ∪ 7 = 7 in the standard encoding of set theory. Because ∪ is the join operation in the natural number lattice (every total order give a lattice structure), yet the lattice structure in ideals of natural number ring is different: the join is LCM and the meet is GCD.

I guess my objection is that the ∪ and ∩ in the set theoretical encoding is rather trivial: the lattice structure in a total order is not terribly informative: join gives the larger element, whereas meet gives the smaller one. Yet the standard encoding of natrual number in category theory (the category generated by one arrow on one object) is slightly more interesting, as composition encodes addition, which is arguably the most interesting opration on natrual numbers.

That being said arguing about encoding of natrual number is not the most informative discussion. but I feel set theory in general is very low level, yet people usually think in more algebraic and high level way, which aligns more closely with category theory.

Counter point, no lawful theory shell let 5 ∪ 7 type check.

If function composition is chaotic, then set intersection is certainly not lawful.

Moxie Marlinspike mentioned why they decided to build a centrualized service: basically, for rapid update to security protocols and less technical burden for backward compatibility.

https://signal.org/blog/the-ecosystem-is-moving/

Indeed there is decentralized E2EE messenger: https://getsession.org/ and it is quite popular (not close to the popularity of signal, though).

Am I too dumb or is this some LLM gibberish that has nothing to do with probablistic programming…

I haven’t used or read about PHP after college, and I am quite interesred in what are the good stuffs that you like now.

It is cool to see how languages evolve.

I heard in many large companies, they would create short artificial downtime for internal services, so that user-facing system never rely on a single internal system for resources and data. This prevents large downtime in user facing system when any internal service is down for a large amount of time.

I think a great use of interns is to make these downtime more organic.

This face is as excited as a cat can get.

It feels very strange to me that any serious citation counter would index ResearchGate, which AFAIK don’t have any check before publishing a preprint. It is basically a more reputable vixra.

But then again citation count, or “impact factor”, are in general quite bad to determine the quality of one’s research, and often can be easily manipulated even through legitimist means: simply publish more mediocre papers.

This is actually gaming the Google Scholar system instead of harmless ones like Willard.

https://www.science.org/content/article/how-easy-it-fudge-your-scientific-rank-meet-larry-world-s-most-cited-cat

Give it another 20 years, PHP might even evolve to catch up with programming languages in the 1980s 😮

TO THE UK GOVERNMENT

In terms of app sandboxing, I don’t think GrapheneOS is doing much to improve privacy, besides sensor and network premission. https://grapheneos.org/features#improved-sandboxing detailed its improvement to sandboxing, but it is more for malicious app, but not for data hungry app, like google product.

In general, an app can only track your behavior within the app besides the premissions you granted (Play service provides an global advertizing ID, which can also be used for tracking, but you can disable it or reset it at any time). You can check these premissions on the premission tab, and the three dots on the top right will show you all premissions.

Only when addition have a inverse operation.

Thank you! This is very helpful.