hocuspocus

It's also more token hungry than similar models, so does it really make a big difference in the end?

With GitHub Copilot pricing, I have found no practical reason to use Gemini 3.8 Flash vs GPT 5.6 Luna. And now I'll probably find no reason to use Gemini 4 when GPT 6.1 Sol is essentially as good and a lot cheaper.

Killing civilians and destroying infrastructure in the name of saving oppressed people; it's quite impressive such low-effort imperialist propaganda still works after Iraq, Afghanistan, Syria, Libya.

If you live in better material conditions than 90% of the world population, you're rich, by definition. This is nearly half the global CO2 footprint.

Even in Europe you can't necessarily bring online a 1200+ MW plant just anywhere, especially in the recent market where some PV production is curtailed in summer and brings the spot rates to zero.

In theory we should start investing in synthetic fuel production, but it's been mostly vaporware so far.

Pretty irrelevant to the last meter, and most ISPs do it for free when you subscribe as they can re-use conduits and sheaths for the legacy copper line, otherwise it's typically a few hours of work.

Where do you think is the last meter if not in the home?

It's not really a big issue. Connecting the last meters of fiber is cheap, either done by default in new constructions, or on demand for older houses. ISP's hardware to run PON is cheap.

Once street level fiber deployment is done, whether market penetration follows or not isn't a huge deal.

The main issue with copper comes from the active infrastructure rather than the physical lines.

I work at an ISP that gets DSLAM equipment from Huawei, we run one of the few G.fast VDSL deployments in the world using this particular hardware. Huawei has stopped manufacturing this product line (the Swiss market is a rounding error ;), which will force us to fully migrate some neighborhoods, only for backup inventory.

Germany being 15 years late, I assume you can still buy similar hardware from AVM though.

And in places where FTTH is available, you pay more for 250/100 Mbps than symmetrical 10 Gbps in Switzerland.

France doesn't have much FTTC, VDSL and HFC exist but are fairly rare.

Orange has started decommissioning its legacy copper network.

No, it's a competing ecosystem, even though both are built on Apache Arrow, and their dataframe APIs might look similar on the surface.

DataFusion is being used as a building block for a growing number of databases and data processing engines in Rust, as it offers the necessary primitives.

Codex, Kiro, Grok Build. Pi has a clone in Rust too.

I assume you don't use Twitter, otherwise you'd have noticed that annoying bots are in the replies.

There are many other clues that clearly make them bots: obvious LLMisms, parroting canned talking points, accounts based in developing countries with nonsensical profile bios, ...

But even without all that, it's very hard to believe that suddenly Twitter attracted all the 4chan-level far-right cesspool organically, and that these people have nothing better to do than replying to any post getting traction from people that should be completely outside their interest circles.

Large-scale offshore wind projects in Europe are now nearly as slow and sometimes more expensive than nuclear.

Solar + batteries are still not enough for countries that have long and grey winters.

I'm fine with this, recent geopolitical events showed that proliferation might have deterred the world's biggest bullies.

You don't even need that. Fast breeder reactors are enough.

Nuclear waste is a nice problem to have, and fairly simple compared to most pollution sources associated with our lifestyle.

No need to worry though, on the current trajectory I doubt our civilization will have the opportunity to deal with it.

You're just supporting my point.

Many European countries do install AC actually, poor ones at that, so it'd be a valid point because yes, it's that easy.

Both sides of the same coin, I'd be surprised if Databricks wasn't looking into getting rid of the JVM entirely.

The data processing space is increasingly tangled with ML and AI, Python on top of C++ and Rust libraries makes sense, keeping the JVM to orchestrate tasks and shuffle data around, not so much.

Right, especially if you were burned by Scala, Kotlin is even less principled, and lately has grown in complexity, partly thanks to re-inventing several Scala features, only worse.

I also talk to pure Java teams, that never considered anything beyond the typical Java/Spring/Hibernate, they won't even have a look at slightly more modern approaches (Quarkus, Panache, build-time annotation processing and native images) despite jumping on new version of Java rather quickly. They couldn't care less about guest JVM languages.

Roughly speaking I believe that developers who enjoy Scala won't go a lesser language, that looks similar on the surface but specifically takes away what makes Scala special. Then people who dislike it will just stick to the most vanilla Java stack, which benefits from being literally one of the most popular choice on Earth.

Of course there are a few people doing Android or KMP, or who think Kotlin is in the right spot on the backend. But usually they aren't coming from Scala.

How's that relevant to people moving away from Scala?

Which moderators exactly? That hasn't been the stance of people in the Scala Center, Typesafe/Lightbend, VirtusLab and SoftwareMill.

ScalaZ is old news and if you went there (on IRC at the time) you knew exactly what you wanted. It doesn't excuse people being assholes, and the Typelevel schism happened partly because of that. Since then people on Gitter and now Discord have been nothing but kind and welcoming to newbies. Same in the ZIO community.

Of course, people building FP libraries will defend their ways. Odersky doesn't build the library ecosystem, and without the FP community Scala 3 would pretty much have killed the language entirely, as nothing else was ready until a few years later. It makes sense that people keeping the main open source communities alive have a bigger presence in the broader community, it's almost tautological.

I haven't seen a huge deal of momentum around Li Haoyi's libraries for instance, they're very much used in their narrow scope, but there's no ecosystem per se. Even his own build tool struggles there despite the technical merits, and he's been advertising Mill non-stop in the past years.

You can't force people to buy into your way of doing things. If there was space for a great non-functional landscape built in Scala then it would exist. Nobody stops you from building and gathering with like-minded people.

Not only it happened at the worst possible time, but some pain points are entirely self-inflicted, like the braceless syntax hurting tooling and fragmenting the community for very little reason.

But in my opinion Databricks implicitly saying "no thanks" for the foreseeable future did a lot of damage here. Databricks didn't even bother bringing their proprietary runtime to Scala 2.13 until last year. Scala 3 is not entirely out of the question but it's taking forever, and at this point users can seriously ask themselves if Spark will actually stay relevant that long.

I try to keep track of teams moving away from Scala at my company and among my friends and acquaintances, and I know a lot of key people in the community.

Scala to Kotlin stories are extremely rare.

Kotlin is almost non-existent in the data processing world (Spark, Flink, Kafka streams...)

Kotlin's presence in the average CRUD web app or more complex SOA doesn't overlap with Scala that much, as someone who enjoyed the paradigms available in Scala would definitely not move to Spring, and frameworks such as Ktor or http4k are pretty niche. I've seen people move to modern Java, Rust (a lot actually), even Go despite the language being at the other end of the spectrum. But Kotlin, not really.

I'm pretty sure most companies don't care nor are aware about the community drama. Arguably Rust is worse, and it doesn't impact its uptake.

Scala was sold through Akka and Spark, solving pretty complex technical challenges that are less relevant nowadays. In fact if you can avoid Akka and Spark, do.

A lot of companies followed the hype without investing in the developer experience. Tech debt hurts more in Scala than in a bigger, slow moving language. After typical turnover, you get teams inheriting unmaintainable codebases. This leaves a bad impression. It can be avoided, Twitter built world-class SOA with Finagle, as well as a complete layer of libraries on top of Hadoop, at a time where the Scala ecosystem was even rougher.

The fact Databricks doesn't care much about the language also did tremendous damage, forcing the community to cross-build against very outdated versions of Scala.

In short, Scala was never a better Java and adopters who thought otherwise invariably got burned.

I wonder how you see 10x more Clojure news.

Scala's usage is stagnant or declining, but Clojure has always been tiny in comparison. It doesn't even register in most rankings (TIOBE, StackOverflow) which is also confirmed by hard numbers e.g. GitHub statistics.

In Scala, .sc is used for worksheets and inconsistently for scripting, see https://scala-cli.virtuslab.org/scripting for instance.

Java LSPs are essentially the same approach, headless Eclipse JDT or NetBeans.

While Metals for Scala was an overhaul from the former Eclipse plugin (and Ensime) plus early experiments embedded in the Scala 3 compiler. And Metals v2 is yet another rewrite.

Metals is architectured in a completely different way, see the BSP: https://build-server-protocol.github.io