Don't forget the thin and tall serif fonts, with one singular italicized word in the title.
196 karma · joined December 14, 2019
Don't forget the thin and tall serif fonts, with one singular italicized word in the title.
DeepSeek: DeepSeek V4 Flash $0.098/$0.196 DeepSeek: DeepSeek V4 Pro $0.435/$0.87
v/s
Anthropic: Claude Haiku $1/$5 Anthropic: Claude Sonnet $3/$15 Anthropic: Claude Opus $5/$25
Many people (me included) sometimes use Opus-class models to plan then swap to Deepseek and/or local models for implementation. The real AI war is in the pricing/performance ratio, and it doesn't look like any of the US-based models are winning on that front.
You would be correct. It uses a wrong tone curve (piecewise-sRGB vs gamma 2.2/2.4) which makes the darks look off: https://github.com/dylanraga/win11hdr-srgb-to-gamma2.2-icm
This was true at least as of 2023-25 when last checked, having been super annoyed by HDR just looking flat out wrong on my PC vs my Mac (which is probably the only OS that gets everything right given their large audience of Creative/visual professionals and mobile experience).
Especially once the mass produced cheap stuff starts being churned out, and there's no cost incentive to go back and fix wrong messaging. USB-IF constantly drops the ball around this ngl, feels like they're a pure scientific community that doesn't think about consumer adoption and UX.
> But HW is at least improving (eg. they added anti-reflective screen option), and SW is very much not.
And I would disagree with the idea that I should be running Linux on my primary machine. As a developer, I've faced enough "death by a thousand cuts" situations from running Linux on my personal router and servers to let it anywhere close to my main computer.
Don't even get me started on the hardware quality of Mac laptop including their stellar trackpads, screens and the smallest details like the quality of the hinge. I can still open my 5 year old Mac with a single finger and the hinge is as solid as the day I bought it.
As someone who's also particular about user experience, Linux always fails at this. If you have good UX, that means you can critically think for what a user wants from a computer, and can determine what should and shouldn’t be prioritized. UX is never a first-class citizen on Linux, and for all the issues with Tahoe, macOS still has enough residual quality left in it to not feel like I'm constantly fighting the operating system.
Simple example: I want HDR on Linux. Should be easy right? Just switch to Plasma under Wayland? Then do a one time config so mpv can play HDR. Oh and no browsers support it so good luck. Games need gamescope and flags to be set.
I want my computer to work, not for me to work as an integration engineer. So I use my Mac and it just works™. So I just let Linux live where I feel it works best, in servers and headless environments.
Used to be an Ollama user. Everything that you cite as benefits for Ollama is what I was drawn to in the first place as well, then moved on to using llama.cpp directly. Apart from being extremely unethical, The issue is that they try to abstract away a bit too much, especially when LLM model quality is highly affected by a bunch of parameters. Hell you can't tell what quant you're downloading. Can you tell at a glance what size of model's downloaded? Can you tell if it's optimized for your arch? Or what Quant?
`ollama pull gemma4`
(Yes, I know you can add parameters etc. but the point stands because this is sold as noob-friendly. If you are going to be adding cli params to tweak this, then just do the same with llama.cpp?)
That became a big issue when Deep Seek R1 came out because everyone and their mother was making TikToks saying that you can run the full fat model without explaining that it was a distill, which Ollama had abstracted away. Running `ollama run deepseek-r1` means nothing when the quality ranges from useless to super good.
> And as someone running at 16gb card, I'm especially curious as to if I'm missing out on better performance?
I'd go so far as to say, I can *GUARANTEE* you're missing out on performance if you are using Ollama, no matter the size of your GPU VRAM. You can get significant improvement if you just run underlying llama.cpp.
Secondly, it's chock full of dark patterns (like the ones above) and anti-open source behavior. For some examples:
1. It mangles GGUF files so other apps can't use them, and you can't access them either without a bunch of work on your end (had to script a way to unmangle these long sha-hashed file names) 2. Ollama conveniently fails contribute improvements back to the original codebase (they don't have to technically thanks to MIT), but they didn't bother assisting llama.cpp in developing multimodal capabilities and features such as iSWA. 3. Any innovations to the do is just piggybacking off of llama.cpp that they try to pass off as their own without contributing back to upstream. When new models come out they post "WIP" publicly while twiddling their thumbs waiting for llama.cpp to do the actual work.
It operates in this weird "middle layer" where it is kind of user friendly but it’s not as user friendly as LM Studio.
After all this, I just couldn't continue using it. If the benefits it provides you are good, then by all means continue.
IMO just finding the most optimal parameters for a models and aliasing them in your cli would be a much better experience ngl, especially now that we have llama-server, a nice webui and hot reloading built into llama.cpp
Point it to a local models folder, and you can firewall the entire app if you feel like it.
Digressing, but the issue with open source software is that most OSS software don't understand UX. UX requires a strong hand and opinionated decision making on whether or not something belongs front-and-center and it's something that developers struggle with. The only counterexample I can think of is Blender and it's a rare exception and sadly not the norm.
LM Studio manages the backend well, hides its complexities and serves as a good front-end for downloading/managing models. Since I download the models to a shared common location, If I don't want to deal with the LM Studio UX, I then easily use the downloaded models with direct llama.cpp, llama-swap and mlx_lm calls.
[1]: https://obdev.at
What I have noticed is to let it load the snipping UI first and start listening for your keystrokes. If you hit escape really quickly, right after you accidentally hit print screen, it will not go away and require you to manually mouse over and hit the X. But if you wait for a second and then hit escape, it almost always registers my keystroke. It's almost like hitting escape really quickly just quits the keyboard listener instead of the actual snipping tool.
If almost every developer-centric forum is constantly complaining about you have enough of a broad sampling of a userbase that there's something rotten underneath is it not? Another ref: See the Reddit thread, also rejoicing at StackOverflow's demise. There's definitely something that they did wrong, and to call it "incorrect" IMO is reductive especially when you have almost every developer practically breaking out champagne at the news.
Communities don’t lose goodwill at that scale by accident.
And full disclosure, I am one of those. I hate StackOverflow with a passion. The holier-than-thou attitude of the moderation playing a major role for sure (and the design that screams QA when they want to be a knowledge-base instead)
> I have been involved in this process for years
Maybe your proximity to the system has made the moderation decisions feel natural when you know the underlying rationale, you can argue that the site is "working as designed", but if the design no longer serves the community it depends on, correctness becomes beside the point, and that's not to say half of what decisions the overzealous moderators make are even correct.
> Or "how is my question a duplicate when actually I asked two questions in one and only one of them is a duplicate?" (n.b. the question is required to be focused in the first place, such that it doesn't clearly break down into two separate issues like that)
Or how about a valid question being closed as a duplicate for a completely different unrelated question? These styles of questions are not uncommon to see: "How do I get red apples?" Closed as a duplicate of "Here's how you make applesauce."
I've had a long shelved project (>8 years now?) where I was working on a solution to doing this from a mobile device but with loads more customization (including image options for different slots), but the cost effectiveness thanks to the PKPass signing as you noticed, put me off to provide it as a public utility as I was a student then. This gives me motivation to revisit it.
NO other (yes I’d die on the hill) open source software has good UX and it’s horrible for adoption by the larger public.
The biggest annoyance I've found is the horrendous battery life on the X-T3. For a long day outside on a trip, I end up going through at least 3 batteries.
The XT-4 is identical to the X-T3 (well, more so than any other x-tn -> x-t(n+1) camera) but fixes a few of the flaws in the X-T3 with massively improved battery life + IBIS which I'd recommend just because a lot of acclaimed lenses these days forgo OIS (ref: many Sigmas for instance), which could be worth it over the long term.
If you are very price sensitive then the X-T3 is still a really good purchase, with nifty features like dual SD slots which make it great to have backups/RAW+JPEG on two cards. Compared to an average photo from a phone, there just isn't much computationally going on in mirrorless cameras so even an x-t1 would be a good purchase.
If you want to shoot photos for the experience rather than getting clinically perfect images, and do not want absolute performance wrt focusing etc., it's definitely at the top IMO; analog with every control having a dedicated physical control (ISO, Shutter Speed and Exposure Compensation and aperture on Fuji lenses). I love it because it's the equivalent of driving an air-cooled Porsche, warts and all.
Adobe is extremely incompetent wrt actually developing usable, clean and efficient software so I can definitely see this as being plausible
They've constantly been downgrading the quality and the polish of the macOS app, just for "cross-platform" feature parity- leading to a subpar experience everywhere (Windows is a whole another can of worms).
So like an intern/new hire asking what they think are basic questions but lead you to realize an edge case you missed?
This is easily remedied by using a very lightweight switch. Many different mechanical switches are available from well under 45gf to actuation force. (Thinkpad and macbook laptop/desktop keyboard range from 60 to 65gf of actuation[1]) You can see a small set of available switches and their force actuation curves here: switches.mx. You should possibly look into getting a switch tester with many types of switches to get the exact feedback force you prefer.
There’s a ton more websites (theremingoat.com) and public github databases for this data.
[1]: https://www.reddit.com/r/MechanicalKeyboards/wiki/ripometer