8
points
virtualritz
Joined 2,912 karma
hacker/founder/syberit
alias: virtualritz at protonmail
meet.hn/city/52.5173885,13.3951309/Berlin
Socials: - github.com/virtualritz - linkedin.com/in/moritzmoeller
---
- virtualritz parentI'm so curious why this was downvoted. Please tell me!
- The two main issues with Lego sets I have are:
The huge amount of specialized parts that are pretty useless for basic building. They're basically adornments.
Prices are insane. I.e. quantity and often quality (as in: how good is the set play-/ construction time wise) is just shite.
I usually buy competitors. Here in Germany Blue e.g. bricks have opened some stores so that's where I take my nephew.
Their sets are much more like the Lego I grew up with. Using more basic parts that exist in creative ways so the specialized/adornment meaning is derived from context, not the part itself/its shape.
Which also requires more imagination from the kids playing with this.
- Unless I miss something I think that this describes box filtering.
It should probably mention that that this is only sufficient for some use cases but not for high quality ones.
E.g. if you were to use this e.g. for rendering font glyphs into something like a static image (or a slow rolling title/credits) you probably want a higher quality filter.
- > I’m waiting for a better terminal experience, personally.
Same! Even colored tabs would go a long way for me.
- 7 USD/day? That's ~200/month -- isn't that just very expensive? I am probably missing something.
E.g. a Terragonlabs subscription is 25/month for 3 concurrent tasks and 50/month for 10.
- What people do not understand is that this really depends on what language you target. So if I write Rust then you sound like an AI hype booster but if I write TS or Python maybe not so much.
From my experience Opus is only good at writing Rust. But it's great at something like TS because the amount of code it has been trained on is probably orders of magnitude bigger for the latter language.
I still use Codex high/xhigh for planning and once the plan is sound I give it to Opus (also planning). That plan I feed back to Codex for sign-off. It takes an average additional 1-2 rounds of this before Opus makes a plan that Codex says _really_ ticks all the boxes of the plan it made itself and which we gave to Opus to start with ...
That tells you something.
Also when Opus is "done" and claims so I let Codex check. Usually it has skipped the last 20% (stubs/todos/logic bugs) so Codex makes a fixup plan that then again goes to through the Codex<->Opus loop of back and forth 2-3 rounds before Codex gives the thumbs up. Only after that has Opus managed to do what the inital plan said that Codex made in the first place.
When I have Opus write TS code (or Python) I do not have to jump through those hoops. Sometimes one round of back and forth is needed but never three, as with Rust.
- TLDR; Not only is the new choice of font unfortunate at best -- the formatting reveals another level of amateurishness so very unbefitting to the ernestness one might assume the sender of the letter wanting to convey. ;)
Basic typography: a paragraph starts with indentation if there is no blank line of any height after the previous one. And if it is not the first paragraph. In short: indent or put vertical space. Never both.
The old TNR version gets this right: if you put blank lines between paragraphs, you don't indent.
Then the date -- dangling god-knows-where, aligned with nothing.
In the old version the only formatting faux-pas is the alignment of 'Sincerely' and if you're picky the outdent of the seal in the top left is a tad much (outside optical axis).
- I just found out that https://annas-archive.li/ is masked by my German internet provider (SIM.de/Drillisch). I usually use a VPN but I had it switched off temp. to watch Fallout (Prime Video won't let you watch through a VPN). Only when I switched Mullvad back on could I open the site.
I didn't know German providers do this.
- As someone who has written C and C++ for 35 years and Rust for a decade now, the last five years professionally, I don't see the point.
Even when I need what C does "well" from the author's pov I can just wrap my code in a big unsafe {} in Rust.
I still get a bunch of things from the language that are not safety- but ergonomics-related and that I would't want to miss anymore.
- > One practical problem I ran into early on is that Google Maps is surprisingly bad at categorising cuisines. A huge share of restaurants are labelled vaguely (“restaurant”, “cafe”, “meal takeaway”)
It's not only that; cuisines are also difficult to label as certain countries simple do not exist for Google when it comes to that.
I recall last year I wanted to change the type of "Alin Gaza Kitchen", my ex (closed now, unfortunately) fav. falafel place in Berlin from the non-descript "Middle-Eastern" to "Palaestinian" category.
I assumed this was available for any country/cuisine, like "German", "Italian" or "Israeli". But "Paleastinian" didn't exist as a category.
- Yeah, and because of this for example Claude Code is down too because the auth goes through CF. F*cking marvelus, the decentralized web ...
- Typing any letter into the search field makes it loose focus. So every letter typed there requires another click first to re-focus.
- There is is, Spielberg spelled wrong as "Speilberg" four (all) times in the article.
Does anyone know why Americans do this regularly, swapping i and e especially in words of German origin?
- I remember a co-student of mine asking me in, probably some time around 1993, I was studying in Hamburg:
Hey, I have a ticket for Björk in Grosse Freiheit this week, I paid 20 bucks, can't go, want it for 15?
Me: Who is Björk? Whatever, sure.
There were maybe 30-50 people on that concert. It was mindblowing; very personal. I didn't know her music at all and ended up 2m away from the small stage. After that I definitely knew who Björk was. :)
- For everyone who is upset about the price tag: this is a wealth-signaling product. Like a Hermès handbag.
Apple is not a luxury brand but it's close enough with some of its products that it can get away with a price like this on a product like that.
- > [...] I'm using Opus 4.1 which is much better but seems to have much lower usage limits than before Sonnet 4.5 was released [...]
Yes, it's down from 40h/week to 3-5h/week on Max plan, effectively. A real bummer. See my comment here [1] regarding [2].
- Curious that this omits Opus.
Opus 4.1 beats Sonnet 4.5 and Codex for me still in any coding tasks. In planning it's slighly behind Codex but only slightly.
Caveat: I do almost exclusively Rust (computer graphics).
- I would really like to understand where that "sometimes" is, nowadays.
RGB just means that color is expressed as a triplet of specific wavelengths. But what is red? And what does red = 1.0 mean w/o context (aka primaries & whitepoint)? What about HDR? What does green = 2.0 convey? Etc.
For context, I worked in VFX production from the 90's to the early 2010's. About 25 years.
And in commercially available VFX-related software, until the early 2000's, mostly, RGB meant non-linear sRGB, unfortunately (or actually: "whatever" would be more true).
And it shows. We have VGX composed in non-linear color space with blown-out, oversaturated colors in highlights, fringes from resulting alpha blending errors, etc. A good compositor can compensate for some of these issues but only so far. If the maths are wrong, stuff will look shitty to some extend. Or as people in VFX say: "I have comments."
After that, SIGGRAPH courses etc. ensured people were developing an understanding on how much this matters.
And after that we had color spaces and learned to do everything in linear. And never looked back.
Games, as always, caught up a decade after. But they, too, did, eventually.
- RGB is a color model[1], not a color space[2].
Note that the headline gets this wrong but the page linked to gets this right.
sRGB or Rec2020 or ACEScg etc. are color spaces with known primaries and a known whitepoint. This is not nit-picky. Almost everyone doing CGI for the first time w/o reading up on some theory beforehand gets this wrong (because of gamma and then premultiplication, usually in that order).
Then there are color models which are also color spaces. CIE XYZ is an example.
- > Surely dithering is good enough to display 10-bit colour on a non-10-bit monitor with good colour.
Indeed, but most software does not do this. It's not on people's minds.
When people hear dithering they think of palette dithering.
People think 8bit/channel is enough; but it is not.
Because it is only 256 levels of any gradient. And because of gamma correction it's practically quite a bit fewer; about 14% (i.e. 220) on a full 256 step gradient that went through an sRGB OETF.
I maintain a Rust crate that solely concerns itself with dithering for such cases as the one you mention.
The readme has a gradient on top that should make it obvious why this matters.
- len() is also returning int instead of uint/uint64 in Go.
I do not use Go but ran into this when I had to write a Go wrapper for some Rust stuff the other day. I was baffled.
- When you say Claude Code, what model do you refer to? CC with Opus still outperforms Codex (gpt-5-codex) for me for anything I do (Rust, computer graphics-related).
However, Anthropic restricted Opus use for Max plan users 10 days or so ago severly (12-fold from 40h/week down to 5h week) [1].
Sonnet is a vastly inferioir model for my use cases (but still frequently writes better Rust code than Codex). So now I use Codex for planning and Sonnet for writing the code. However, I usually need about 3--5 loops with Codex reviewing, Sonnet fixing, rinse & repeat.
Before I could use one-shot Opus and review myself directly, and do one polish run following my review (also via Opus). That was possible from June--mid October but no more.
- Does anyone have a clone of the GH repo shared in the original article? It seems to have disappeared from GH.
- Indeed. In English it is an em-dash. Not brackeded by spaces, although you should bracket by thin spaces to make sure you get some separation, optically. And line breaks in apps that are unaware this is a separator.
In many other languages ideas are separated by en-dashes (surrounded by spaces).
In the case at hand the use of en-dashes in English text instead of the correct em-dashes could also be a sign of a non-native speaker caring about micro-typography and just doing the wrong thing. :)
The LLM conclusion doesn't seem well supported either way.
- These are 'en' dashes. Myself and and other people who care about (micro)typography have been using them since forever.
Presumably the author is one of them. Or they simply use a text editing or blogging software that takes care of it.
E.g. Markdown with smarty-pants feature turned on generates them automatically from '--'. 'Em' dashes require '---'.
Coincidentally Rust's `cargo doc` does this for you -- just for example.
The conclusion that a text containing such micro-typographic niceties must be LLM-generated is a fallacy thusly.
Your other 'evidence' sounds like an interpretation to me. Maybe you should quote the sections you mean?
Otherwise your critique seems superficially limited to form, not contents -- an ad-hominem in disguise one may be tempted to conclude.
- I'm doing computer graphics code, 2D, 3D, all CPU (not GPU) and VFX related. It's a niche topic for sure. Often relate to research papers that come without code.
I.e. I can tell from the generated code on this vs. other 'topics' that the model has not seen much or any "prior art".
- > Everyone believes their task requires no less than Opus it seems after all.
I have solid evidence that it does. I have been using Opus daily, locally and on Terragonlabs for Rust work since June (on Max plan) and now, since a bit more than a week, being forced to use Sonnet 4.5 most of the time. Because of [1] (see also my comments there, same handle as HN).
Letting Sonnet do tasks on Terry, unsupervised is kinda useless as the fixes I have to do afterwards eat the time I saved giving it the task in the first place.
TLDR; Sonnet 4.5 sucks, compared to Opus 4.1. At least for the type of work I do.
Because of the recent Opus use restrictions Anthropic introduced on Max I use Codex to planning/eval/back and forth (detailed) and then Sonnet for writing code. And then Opus for the small ~5h window each week to "fix" what Sonnet wrote.
I.e. turn its code from something that compiles and passes tests, mostly, into canonical, DRY, good Rust code that passes all tests.
Also: for simpler tasks Opus-generated Rust code felt like I needed to glance at it when reviewing. Sonnet-generated Rust code requires line-by-line full-focus checking as a matter of fact.
- I tried both and they're mostly useless for what I do (Rust, computer graphics). But so is Sonnet 4.5.
Best "Opus replacement" so far is using Codex for doing the thinking and writing the code which Sonnet 4.5 then goes over to make it actually compile ...
But that also means I now need to manually coordinate two models/agentic tools whereas before this was all mostly running automatic and all I did was specs, code reviews/write tests.
What this now comes down to, 10 USD/h for Opus 4.1 (20h of Opus/month @ 200 USD) is not worth it for me.
It seems https://jules.google.com/ is currently free? Has anyone tried this?
- See also this reddit thread where Anthropic have confirmed this policy change:
https://www.reddit.com/r/ClaudeAI/comments/1nvnafs/update_on...