> Even my moderate view, though, is premised on the assumption that A.I. will continue to advance up the steep part of the sigmoid curve for a while before hitting one or another physical constraint that creates a new inflection point and slows advances further.
I feel like we've pretty much already hit a fundamental barrier in compute that is unlikely to be overcome in the near future barring a profound, novel algorithmic approach or an entirely novel computing model.
Are there any events in the last, maybe 6 months, that seem to be still on the steep part of the sigmoid curve? I'm probably not well informed, but I can't think of any. GPT5 launch sure does not feel like a steep advance. What else was there?
I think any comment about an AI bubble needs to start by defining who the players are and how it affects them differently.
AI foundries, Nvidia, the hyperscalers, enterprise buyers of AI, consumers, the US, China, the rest of the world, startups, investors, FOSS, students, teachers, coders, lawyers, publishers, artists... each stand to win or lose in profoundly different ways.
Bolsonaro's policy choices regarding the Amazon likely meaningfully contributed to this issue by allowing "ideal conditions" to accumulate in the years that preceded. If anything, this helps confirm it rather than absolve him.
Naturally other issues like El Niño and factors outside of Brazil also play a role, but it stands to reason this would not have been as tragic had he not been elected president.
Nah. Time to stop blaming him. It's been three years, plenty of time for Lula to do something if he actually gave a shit. He's too busy to actually run the country though. He's got far more important things to do such as finding innovative new ways to tax us and persecuting Bolsonaro until his death.
This is a nonissue. I've written several lengthy documents mixing up all sorts of pieces in Typst over a year, created my own template, modules, functions...
I don't think I do. Are you going to run a bus every 15 minutes down a road that would have one passenger an hour? Mass transit isn't viable at the density of the suburbs but building higher density there is banned.
We've incentivized cities to develop around highways and the automobile infrastructure instead of building them for mass transit. You need cars because we build for cars.
It's not that we've incentivized cities to develop around highways, it's that we've prohibited them from doing anything other than that.
Zoning boards put a tiny little strip of commercial and high density residential in the downtown and then require the whole rest of the map to be single-family homes. At that point it doesn't even matter what the downtown actually looks like, people are still going to be in cars because it's the only way to get there from the suburbs.
These zoning board decisions were made largely to accommodate cars. For example, in many places, we can't have dense urban housing or commercial unless the developer pays to park all of the cars associated with the new development (so the cars don't consume public street parking). But this means we end up surrounding buildings with these giant parking lots which creates more space between each building, putting downward pressure on walking/transit and upward pressure on driving. This also means you need more lanes to accommodate the cars (the additional lanes also create more space between buildings and make pedestrian traffic considerably less desirable, putting more upward pressure on driving).
Tangentially, the additional length and width of roads as well as the traffic lights all constitute an increase in infrastructure costs while also reducing the amount of revenue generated per unit space (because so much more of the space is for streets and parking).
The space between buildings thing is a red herring. If you want an area full of tall buildings, there must be a significant amount of space between them to let in light and fresh air. You could hypothetically use that space for greenery or something instead of lanes and parking but you can't get rid of it and use it to increase density. Moreover, it isn't actually a density limit anyway because you can make the buildings taller instead of wider, and you can build a parking garage under the building rather than beside it.
The real thing minimum parking requirements do is increase cost, because building parking floors costs money. But that isn't nearly as much as the cost increase from zoning most of the map exclusively for single family homes, because that's the thing that makes the land expensive, and on top of that requires you to use 15+ story buildings in the limited area that allows them when you could have the same average density by using 3-5 story buildings over a wider area.
Moreover, you can't put the cart before the horse. If people currently live in the suburbs and arrive in cars, you can't expect them to walk before you allow anyone to build them housing within walking distance.
> The space between buildings thing is a red herring. If you want an area full of tall buildings, there must be a significant amount of space between them to let in light and fresh air.
First of all, I don't think anyone's goal is "an area full of tall buildings"; that's certainly not what I mean by "density" (although it is _one kind_ of density). Secondly, even in urban areas full of tall buildings, there's frequently much less space between buildings than a CostCo parking lot.
> Moreover, it isn't actually a density limit anyway because you can make the buildings taller instead of wider, and you can build a parking garage under the building rather than beside it.
Building vertically is expensive, and in many places land is cheaper, so it's easier to meet the legal requirement by surrounding the building with pavement than it is to build a parking garage beneath the structure. This is why you rarely see a Walmart with an underground parking garage (and when you do, it's usually in a dense city with more lax parking regulations).
> Moreover, you can't put the cart before the horse. If people currently live in the suburbs and arrive in cars, you can't expect them to walk before you allow anyone to build them housing within walking distance.
I think you're confused about what is being advocated. No one is suggesting we make everyone walk to work. I don't think that's a realistic outcome, and probably not a desirable one for many people (who wants to work close to a factory, airport, etc)? More importantly, relaxing parking requirements on developers doesn't make the existing parking lots go away, so it doesn't really affect the current crop of commuters; it just means that future suburban commuters will lean more on public transit to get to work.
I do like Swift but it also suffers a bit from identity crisis. The compiler experience is quite disappointing too—I found myself helping the compiler more than the compiler helped me.
I've since moved to Rust and have not looked back. Importantly, rust-analyzer runs circles around the Swift VSCode plugin (or Xcode for that matter)
The identitiy is clear, it is a language first and foremost for Apple ecosystems.
It also needs to target GNU/Linux, because Apple got rid of their server offerings, thus anyone doing server code for applications on the Apple ecosystem, that wants to stay in a single language needs to be able to write such software on GNU/Linux with Swift.
Windows well, since they have the open source story, it kind of falls from there as complement.
On the revamped website they are quite clear about the identity.
Cloud Services, CLI and Embedded as the main targets for the open source version.
I'm the same but opposite, I like Rust but find myself using Swift most of the time. They sort of do the same thing but coming from opposite directions. Can't go wrong with either imo.
I really do wanna try Kotlin at some point as well. Rust, Kotlin, and Swift feel like the future of languages to me.
I would say yes. I have experience teaching Rust to ~20 yo students of Java, and they are able to be productive in Rust within a semester. Your median Java and C# dev should be able to use Rust. Dunno about Python devs.
But I want a fast on-ramp, quick iterations and clean looking code. (and went with Kotlin because of that -- I like Rust more myself, but I have a business to run, so tradeoffs)
I feel like we've pretty much already hit a fundamental barrier in compute that is unlikely to be overcome in the near future barring a profound, novel algorithmic approach or an entirely novel computing model.
reply