Quinten Kock @colonelphantom.bsky.social · 27/03/2026I just consider the reasoning and I agree with both: Haskell is big on partial application. It makes sense to use this with the error type to get a specialized result type. Rust's ordering is more natural though. I do hate the name "Either" though. It makes both options sound too 'equal'. 000
Quinten Kock @colonelphantom.bsky.social · 25/03/2026I do think that the "tech" sector in that sense is more driven by things like VC money, which seems disproportionally affected by ZIRP. 222
Quinten Kock @colonelphantom.bsky.social · 14/03/2026why do you prefer lcds over oleds? all I can think of is burn-in and price 120
Quinten Kock @colonelphantom.bsky.social · 04/03/2026Mozilla has been quite decent as well in my opinion (apart from the chatbot sidebar), though I don't use any Apple products so can't really compare there. Firefox now has an "AI killswitch" blog.mozilla.org/en/firefox/a... and most features are both local and useful (like translate or tab grouping).blog.mozilla.orgAI controls are coming to Firefox | The Mozilla BlogAI is changing the web, and people want very different things from it. We’ve heard from many who want nothing to do with AI. We’ve also heard from 040
Quinten Kock @colonelphantom.bsky.social · 04/03/2026Het enige wat van mij niet zo had gehoeven is dat er een zijbalk is om ChatGPT o.i.d. in te draaien. Toen ze dat aankondigden maakte ik me wel even zorgen dat Firefox Fireslop zou worden, maar het lijkt alleszins mee te vallen. 010
Quinten Kock @colonelphantom.bsky.social · 04/03/2026Mozilla doet het op dit gebied ook heel redelijk: ze werken met name aan lokale AI features, die bovendien daadwerkelijk nuttig zijn (denk aan het vertalen van webpagina's of je tabjes groeperen). Plus sinds de huidige versie is er een AI-killswitch als je dat niet wil. 100
Reposted by Quinten Kockklieret.bsky.social @klieret.bsky.social · 04/03/2026Bloated patches: LM generated solutions of SWE-bench tasks are consistently longer than human-written gold solutions (and it's not just comments) 🧵 111
Reposted by Quinten KockDavid Buchanan @retr0.id · 18/11/2025• web browser (noun) a fork of chromium • text editor (noun) a fork of vscode 1344367
Quinten Kock @colonelphantom.bsky.social · 13/11/2025> Engelen: ‘Die anekdote wilde ik net gaan vertellen, maar het is helemaal goed joh.’ > Thieme: ‘O nee, vertel, vertel.’ > Engelen: ‘Het is wel een beetje mijn anekdote, dus je moet er gewoon met je tengels van afblijven.’ Jezus wat tenenkrommend 030
Quinten Kock @colonelphantom.bsky.social · 29/10/2025True, but it was also expected that the VVD would get decimated these elections, which did not happen. So it could very well embolden them. I guess we'll see, but either way I still firmly believe negotiations will be extremely difficult. Our last three(!) formations were over half a year long. 140
Quinten Kock @colonelphantom.bsky.social · 29/10/2025VVD-leader Yesilgoz has however said she did not want to form a government with GL-PvdA. I also have a hard time seeing D66 together with JA21, so I feel like we're going to have a very difficult formation, as is tradition at this point. 170
Quinten Kock @colonelphantom.bsky.social · 24/01/2025At least I'm pretty sure Clippy didn't cost extra? 000
Quinten Kock @colonelphantom.bsky.social · 24/01/2025how haskell-specific is this problem? it feels like something most type systems would struggle with, since you need to have the forall somewhere. 200
Quinten Kock @colonelphantom.bsky.social · 11/01/2025When you're trying to do some AVX programming using ISPC but you're on a Zen4 machine.. Even doing this serialized is faster lmfao 010
Quinten Kock @colonelphantom.bsky.social · 11/01/2025Make your own text editor 🥸 also, I personally quite like Kate, but I think it's best if you use KDE, and there is no plugin ecosystem to speak of, nor do I think making one is easy (pretty sure it needs to be done in C++...) 010
Reposted by Quinten Kockgabby @fullmoon.id · 09/01/2025Another common mistake I see with LLMs is sticking to prose inputs and outputs for the model Most users are actually pretty allergic to both typing out and reading large blocks of text. It's often better to structure inputs and outputs to be something other than sentences/paragraphs 291
Reposted by Quinten KockBetter Things Are Possible @internethippo.bsky.social · 03/12/2024(pulling up the ladder behind me) This is an opportunity for you to succeed through hard work and determination 283326395
Quinten Kock @colonelphantom.bsky.social · 01/12/2024I just started doing #adventofcode in #Haskell, but I've already done that for like the last three years, so maybe it's time for something else? Does anyone here have any fun suggestions? 020
Reposted by Quinten KockCori 🦝 (DemShenaniganss) #TDST⚡️ @demshenaniganss.bsky.social · 18/11/2024QUICK REMINDER NEWSKIES: Pronouns are SUPER easy to add Just pop over here, sub, and then go to posts to select yours so they pop up as a profile label and you have more room in your bio!! bsky.app/profile/pron...bsky.app 65238
Quinten Kock @colonelphantom.bsky.social · 28/11/2024Seems like it does, it probably defines foldMap and then goes left-through-right: 120
Quinten Kock @colonelphantom.bsky.social · 28/11/2024Fedora offers an atomic Sway spin: fedoraproject.org/atomic-deskt... Might be a nice way if you do want to give i3-style tiling a try!fedoraproject.orgFedora Sway AtomicFedora Sway Atomic provides the popular Sway tiling window manager in an atomic fashion. It makes Sway accessible and appealing to both novices and advanced users who prefer not to use a mouse, touchp... 010
Reposted by Quinten KockDr. Casey Fiesler @cfiesler.bsky.social · 27/11/2024Hi, so I've spent the past almost-decade studying research uses of public social media data, like e.g. ML researchers using content from Twitter, Reddit, and Mastodon. Anyway, buckle up this is about to be a VERY long thread with lots of thoughts and links to papers. 🧵 59959450
Quinten Kock @colonelphantom.bsky.social · 28/11/2024I knew you could derive Functor, but I didn't know it also worked for Foldable. True 🤯 moment for me, tbh. I wonder though, will Foldable behave differently if Node is defined as `Node (Tree a) a (Tree a)`? 220
Quinten Kock @colonelphantom.bsky.social · 28/11/2024Unicode is annoying because it's hard to type. And Haskell typically is pretty rich in symbols (I mean, what the hell is <*>?) so it makes sense to me that they don't use a keyword like fn. Rust's |x| would work, I guess, but has the same problems. Just -> would make the syntax more complex IMO. 110
Quinten Kock @colonelphantom.bsky.social · 28/11/2024Firefox does not share my posts publicly or anything like that. The fact here is that people are mad about the JSON file being distributed publicly on HF. The file hosting does not federate with Bsky so I don't think it's covered, but just saying "JSON file" is a reductio ad absurdum. 100
Quinten Kock @colonelphantom.bsky.social · 28/11/2024The dataset does not include the post author name in a human readable way. As such, searching your or someone else's username is completely pointless. 120
Quinten Kock @colonelphantom.bsky.social · 28/11/2024Alpindale uploaded the file to Huggingface, a SERVICE to share AI content. Doing that provides a SERVICE to AI developers. 100
Quinten Kock @colonelphantom.bsky.social · 26/11/2024So I guess the strictness analyzer succeeds here? That's pretty neat! Still not sure how I feel about relying on a compiler optimization to get good asymptotic memory usage though, especially since it may also require inlining? Guess I'll be trying a few things today! 040
Quinten Kock @colonelphantom.bsky.social · 25/11/2024I'm also curious to see if the i3 lineup will get a cluster of E-cores, especially now that hyperthreading is gone. Making a 4c4t CPU seems kind of not-the-move even for a budget chip, since all i3's since 10th gen have been 4c8t. 4p+4e 8t would be fine and probably a really nice MT uplift too. 000
Quinten Kock @colonelphantom.bsky.social · 25/11/2024I overall quite like laziness, but IMO oversized thunks are a significant problem. For example, folds require explicit strictness to implement a basic reduction like sum without requiring linear (stack) space. That said, I've never tried if the strictness analyzer helps in their case. Does it? 120
Quinten Kock @colonelphantom.bsky.social · 25/11/2024What's wrong with how Tumblr does it? The really long chains, or is it something else? 110
Quinten Kock @colonelphantom.bsky.social · 20/11/2024Yep, and I can't imagine a $3000 PC having older parts either. 000
Quinten Kock @colonelphantom.bsky.social · 20/11/2024Lunar Lake is 200V specifically (not 200H or 200U or 200 or 200K). As for the 13th/14th gen issues, those seemed to be hardware degradation causing actual errors (such as BSODs), not things becoming laggy. The lag is probably just a software/driver issue and not harmful for the HW. 100
Quinten Kock @colonelphantom.bsky.social · 19/11/2024Can't you remove the faulty stick and just use the system with less RAM? It would cause a performance hit but IMO better that than an unreliable system. 010
Quinten Kock @colonelphantom.bsky.social · 19/11/2024Does Turing not have RTX HDR support? According to PCGamer it should work on any RTX card. 000
Quinten Kock @colonelphantom.bsky.social · 19/11/2024I believe that Intel's XeSS is much closer to DLSS in image quality than FSR, since it's also based on machine learning. Especially true if running on XMX-capable Intel hardware (Arc dGPUs primarily). I've heard rumors that the next FSR version is much better, though. 000
Quinten Kock @colonelphantom.bsky.social · 16/11/2024The 285K is great in productivity and the gaming performance hit is negligible in most cases, but there are some massive regressions too. Hardware Unboxed found the 285K to be 20% slower than the 14900K in Cyberpunk 2077. That puts it behind the 12600K, while using more power! 010
Quinten Kock @colonelphantom.bsky.social · 07/11/2024I still have no idea how they managed to make the 285K slower(!!!) than the 14900K in gaming. Something must've went terribly wrong, maybe in the tile-based design? 100
Quinten Kock @colonelphantom.bsky.social · 07/11/2024AMD in 2011: Our flagship CPU has.... 8 cores! (Of course you can get into the whole cores vs. modules lawsuit, but the point stands). Intel was worse thought; their 2019 flagship (10900K) had 10 cores and their 2021 flagship (11900K) only 8! 000