Simon Felix

This is probably my final message on Twitter/X - I'm switching to Mastodon. See you on the other side.

Why are experts so often wrong? Because we have lost track of what an expert is. An expert is someone who has a deep experience with a given set of problems and a track record of getting things mostly right. An experienced plumber is likely an expert at plumbing. An expert is not someone who has studied a topic deeply. That’s a scholar not an expert. What expert would be relevant for predicting the future of AI? People have consistently predicted AI progress properly. If such people exist, they are uncommon. In fact, for many problems, there are few relevant experts. We just don’t know. Often that’s the best answer.

@STeplyakov I consider async/parallel two very distinct topics. For desktop apps, I prefer to run purely async & non-blocking code with unlimited concurrency on a single UI thread, and perform parallel work on a bounded thread pool. Best of both worlds.

Blogged - "You probably should stop using a custom TaskScheduler": sergeyteplyakov.github.io/Blog/csharp/20… I recently posted about the concurrency control and some folks mentioned to use TaskScheduler for that. This is my reasons to avoid that.

@DiademGames I, a small part of the vast emptiness of the internet, agree. Main point is probably whether you're code- or art-/design-driven. Case in point: Playdead switched from custom to Unity, and like it.

@deiruch Better late than never ;) diademgames.com/blog/why_custo… #gamedevs #indiegaming

@thebookisclosed @_h0x0d_ What is Explorer doing for several hundred milliseconds after double-clicking a folder? Why does the volume slider visibly re-enact all volume changes since it was last shown?

Are there any questions you have regarding how particular Windows features / functionality works under the hood? 🧐 I'd be happy to explain some more opaque or odd parts of the system 🔍 Privacy related questions also count, but looking into those might take longer than others ⌛

MatMul-free LLMs appear to be significantly more energy efficient: arxiv.org/pdf/2406.02528. 10x less memory, 5x faster. Interesting!

@STeplyakov And PARALLEL.ForEachAsync shouldn't be about parallelism? Huh... :)

@deiruch PLINQ is about parallelism, not asynchrony;)

Hey #dotnet people. Does anyone know why there is no `Parallel.ForEachAsync` overload that returns `Task<TResult[]>`? It's really annoying that all the overloads are returning just a `Task` :(

@STeplyakov Are there disadvantages to using PLINQ? Why not just use that?

@deiruch That's annoying. I posted just Yesterday when I noticed that. We have a custom version of ForEachAsync that "projects" the results.

@STeplyakov Btw: Not a single Parallel.For/ForEach function collects & returns results.

@STeplyakov I use LINQ when I want to collect results. I don't think I'd want a by-default-array-allocating Parallel.ForEachAsync.

@cybersandwich @karpi Die Zeitung stellts doch ziemlich neutral dar, oder etwa nicht?

Warum wird das immer als etwas Schlechtes dargestellt? Mehrsprachig aufzuwachsen, war etwas vom Besten, was mir passieren konnte. Wir haben zuhause nur bosnisch gesprochen. Schönste Erinnerung: Meine Mutter übt mit mir (11) Französischdiktat, ohne ein einziges Wort zu verstehen.

@dcuthbert @LargeCardinal Nice paper, but the QUBO solving and Q hardware benchmark is... uhm... "not convincing". What was the classical exact algorithm? How was it implemented? Why compare a heuristic with an exact solver? How about comparing with state of the art: Gurobi, Hexaly, ...? Etc.

Patch ALL teh things we constantly tell CISOs and CIOs. Thing is, let's be honest with each other right? we can't and this graph is telling. Patching is a pain, we get it and we do need to revolutionise the approach. Two years ago, @LargeCardinal wrote a phenomenal paper

@karpi Mit welchem Ziel? Bessere Regulierung? Fair trAIne? Nur noch Produkte von Unternehmen mit freilaufenden GPUs kaufen? nVidia Maxwell Havelaar?

@deiruch Na Aktivismus, Politik, Unternehmertum, Kunst. All die öffentlichen Sachen halt.

"Stell Dir vor es ist Hype und keiner macht mit." Heute in der NZZ am Sonntag, mein Text übers Verschlafen und Verweigern der KI-Transformation. epaper.nzz.ch/article/8/8/20…

"Stop what you're doing, you're late for a meeting I've just decided starts immediately. Here, lemme set off a siren in your home until you join." Phonecalls are wild.

It's only consciousness if it's from the Sapiens species of the Homo genus, otherwise it's just glorified pattern matching

@andrewlocknet There's no need to call remove(). Should run faster without it. It's no problem to process the same city twice.

Blogged: Implementing Dijkstra's algorithm for finding the shortest path between two nodes using PriorityQueue in .NET 9 buff.ly/3VuN2xJ In this post I describe Dijkstra's algorithm and show you can implement it with the new PriorityQueue.Remove method in .NET 9 #dotnet

It's a bit of a rant but it has a link to a really useful article from Herb Sutter in it and some of my opinions might be illuminating :D ricomariani.medium.com/c-for-security…

@STeplyakov Hmm. Other impls beat the native impl for common cases: github.com/NetFabric/Linq…

@deiruch Any sufficiently complicated abstraction is leaky. So it always going to be performance penalty for such high level thing like LINQ. Plus, LINQ deals with IEnumerable, which is a very weakly defined abstraction that can be a generator or a collection.

Going to give an internal talk "Demystifying LINQ Performance". Curious how many people would be interestring to see something like this in public?