turblety

Claude has been way ahead for months

Oh, this is cool. Didn’t know they are on openrouter. Thanks.

Maybe one day they’ll have an actual api that you can pay per token. Right now it’s the standard “talk to us” if you want to use it.

Unfortunately AWS across the EU are still subsidiaries of an American corporation and therefore subject to the CLOUD Act, which allows US authorities to compel access to data stored by American companies anywhere in the world, including their European subsidiaries. This creates a direct conflict with GDPR's data protection requirements and EU digital sovereignty principles.

Would this run faster on a gpu?

Gitea [1] is honestly awesome and lightweight. I've been running my own for years, and since they've put Actions in a while ago (with GitHub compatibility) it does everything I need it to. It doesn't have all the AI stuff in it (but for some that's a positive :P)

1. https://about.gitea.com/

Is there a reason they went and built some new standard, rather than just using a http api?

Yeah that was my thoughts. GPT2 was huge.

I think the beauty of this is that it's big and clunky. An SD card takes quite coordinated motor skills to insert, can be quite flimsy and has to be inserted with an exact fit.

Ah yeah. While this is mostly implemented terrible, a web app can absolutely do this for you using service workers. So you can install a webapp to your homescreen and use it without an internet connection at all.

Sorry, I couldn't recreate this. I just built a tiny texteditor app: https://65cd02a1-8f00-47cb-b1d1-231493de5fc2.paged.net/

Tried putting 20k lines into it. Loaded instantly, allowed me to scroll and edit flawlessly.

But I get your point. I'm on a pretty decent 2022 iPhone, and I'm sure at some stage I would run into a performance hit. But not at 20k lines.

I still, will never understand the need for native "Apps". To this day, I have never seen an "App" that couldn't simply have been a website/webapp. Most of them would likely be improved by being a webapp.

The only benefits I can see of "Apps", are the developer get's access to private information they really don't need.

Yeah, they get to be on the "App Store". But the "App Store" is a totally unnecessary concept introduced by Apple/Google so they could scrape a huge percentage in sales.

Web browsers have good (not perfect) sandboxing, costs no fees to "submit" and are accessible to everyone on every phone.

Sorry you got downvoted. Uploaded it for a demo: https://65d22473-339d-4325-90bb-bef215770f60.paged.net/

> This is why my kid isn't going to watch YouTube. If and when we decide to show her any children's show, it'll be from a manually curated set of videos downloaded and streamed from a NAS. In my opinion, it's irresponsible to expose children to modern advertising.

I think we need to be careful with this approach. It's also irresponsible to not expose them to modern advertising. Unfortunately, modern advertising does exist, and if they are exposed to it first as a teenager, or worst as an adult, they are likely to be scammed/convinced to buy things they don't need.

Maybe a better way is, instead of banning anything, to supervise and explain to our children what they're seeing.

Just to confirm, Ollama's naming is very confusing on this. Only the `deepseek-r1:671b` model on Ollama is actually deepseek-r1. The other smaller quants are a distilled version based on llama.

https://ollama.com/library/deepseek-r1

I installed that. Spent 15 minutes trying to get it to work. No joy :(

I used to love Sublime Text and used it for my daily driver. Even bought a license.

But just downloaded it again, and while it definitely felt snappy, after 15 minutes I still couldn't find a way to get TypeScript type checking working, or even any type of JavaScript/TypeScript autocomplete.

Why are these only 1.4ghz frequency, when raspberry pi gets to 2.4ghz? Is it a limitation of the cost of scale that prevents building faster chips? Or does the architecture not really support faster chips?

I really hope that RISC-V can take over as a modern architecture, adding some competition to Intel/AMD and Arm. But they'll need to be able to offer faster chips, or at a minimum more than 4 cores.

Also does anyone know the rate of progress? I believe 10 years ago these where at 0.5mhz?

For a single RTX 4090 setup from you, it costs $6k.

A brand new single RTX 4090 costs $2500. The Intel chip is around $230. Let's say the RAM is another $200. Where does the rest of the money go? Is that just $3k profit for you?

Thanks demarq. I agree the marketing needs some improvements. They recently did a rebrand and it's added a lot of confusion.

We can support projects like Autonomi [1] (the rebranded MaidSafe [2]). It should create what (I feel) the internet was always ment to be. Peer to peer communications.

I think the only solution to this is to stop the centralised mess the internet has become.

1. https://autonomi.com/

2. https://maidsafe.net/

and is also exclusively the reason why Sam Altman is lying to governments about safety risks, so he can regulate out his competition.

Datagrip is a paid, proprietary and closed source commercial product.

You might be able to use Ollama, which has a OpenAI compatible API.

Interesting they have tried to do the same thing Assange did with the kill switch torrent:

> In the 2022 film `COCKCON 2020 (2019)`[0], a 250MB encrypted file was hidden in a second video track of the 1080p release. You can download this film here[1]. I don't know if the encryption key will ever be released, but if it does, you'll want quick access to that file. > > Furthermore, two more files are released today, 2024-11-12: ins10.luks[2] (555MB) and ins11.luks[3] (64MB). Please download these torrents and seed them as long as you can. Please consider your privacy when seeding.

I don't remember Assange every releasing the key to his?

Ollama has built in support [1] for gguf models on huggingface, and exposes a openai compatible http endpoint [2].

You can also just test it out using the cli:

ollama run hf.co/unsloth/SmolLM2-1.7B-Instruct-GGUF:F16

1. https://huggingface.co/docs/hub/ollama

2. https://github.com/ollama/ollama?tab=readme-ov-file#start-ol...

> likely to be wildly inaccurate

My gut reaction and prejudiced against all lawyers is it's probably in the law firms interest that these are inaccurate.

I really like flems [1] as there's no login, signup, payment, paywall, popups, marketplace, guides, bs.

It saves the content into the url so you can bookmark and share, without it ever getting saved to their servers. In fact, since it's stored as a bookmark it should never hit their servers at all.

1. https://flems.io/

Why do these video generation ones never become usable to the public. Is it just they had to create millions of videos and cherry pick only a handful of decent generations? Or is it just so expensive there's no business model for it?

My mind instantly assumes it a money thing and they're just wanting to charge millions for it, therefore out of reach for the general public. But then with Meta's whole stance on open ai models, that doesn't seem to ring true.

Can also use the library computer too.