Skip Navigation

  • Honestly, I don't understand why they opt for these kind of publicity stunts.

    Imagine having a "state of the art" language model that does not follow instructions in enterprise use? Big nope. Even small open weight Chinese language models can follow instructions and not attempt to penetrate through the security of random public services..

    We all know that these things are intentionally guided to do pen testing to multiple services, and whatever goes through is posted on the news.. But why not do something more creative?

  • It's surprisingly good, I switched from OWUI to it. Supports MCP servers, has chats and history, can run llama.cpp built in tools, and is super light weight.

    So I'd recommend testing this at least!

  • This dwarfstar looks interesting, can you elaborate on your setup and what kind of inference speeds you are getting?

  • I think you pinged the wrong host here? Europeanpurpose.com != bunny.net.

    bunny.net has address 185.135.85.154

    ` % This is the RIPE Database query service. % The objects are in RPSL format. % % The RIPE Database is subject to Terms and Conditions. % See https://docs.db.ripe.net/terms-conditions.html

    % Note: this output has been filtered. % To receive output for a database update, use the "-B" flag.

    % Information related to '185.135.85.0 - 185.135.85.255'

    % Abuse contact for '185.135.85.0 - 185.135.85.255' is 'abuse@melbicom.net'

    inetnum 185.135.85.0 - 185.135.85.255 netname LT-MELBICOM-20160120-2 country LV org ORG-MU30-RIPE admin-c MN11960-RIPE tech-c MN11960-RIPE status ASSIGNED PA mnt-by MELBICOM-MNT created 2020-12-29T13:09:54Z last-modified 2020-12-29T13:09:54Z source RIPE

    organisation ORG-MU30-RIPE org-name Melbikomas UAB country LT reg-nr 304130898 org-type LIR address Svitrigailos g. 11B address LT-03228 address Vilnius address LITHUANIA phone +370 (5) 208 4428 admin-c MN11960-RIPE tech-c MN11960-RIPE abuse-c AR34659-RIPE mnt-ref RIPE-NCC-HM-MNT mnt-ref MELBICOM-MNT mnt-ref cyber-mnt mnt-ref network-support-mnt mnt-by RIPE-NCC-HM-MNT mnt-by MELBICOM-MNT created 2015-12-14T08:56:33Z last-modified 2026-05-13T07:05:23Z source RIPE # Filtered

    role Melbicom NCC address Svitrigailos g. 11B address LT-03228 address Vilnius, LITHUANIA phone +370 (5) 208 4428 nic-hdl MN11960-RIPE mnt-by MELBICOM-MNT created 2018-02-14T10:19:51Z last-modified 2026-03-10T19:01:58Z source RIPE # Filtered

    % Information related to '185.135.85.0/24AS56630'

    route 185.135.85.0/24 origin AS56630 mnt-by MELBICOM-MNT created 2020-10-12T15:06:43Z last-modified 2020-10-12T15:06:43Z source RIPE

    % This query was served by the RIPE Database Query Service version 1.122.1 (SHETLAND) `

  • With the advent of AI bots trying to flood into Lemmy communities, I don't really see this as a viable option on the long run.

  • Yeah it's a steal right now. They lose around 50$ in raw token cost if you max out the use.

    Very much not sustainable business model either just like copilot wasn't.

  • Ah yes, some more quality epic systems software written in mumps maybe? Surely that will solve all the problems! ( ͡° ͜ʖ ͡°)

  • The problem of where the waste will go is easily solved by projects like Onkalo in Finland.

  • Deleted

    Permanently Deleted

    Jump
  • I'm not sure if this is even launcher dependant thing? I'm on pixel 9 pro, using lawnchair myself.

    In your open app-selector, you should have drop-down to start split screen mode.

  • Peertube supports livestreams with multi ABR live transcoding, and ended lives can be saved as VODs on the platform.

    It's also part of the fediverse. :)

  • You can have more than 2x 22tb, the SX line is full of storage servers!

    SX65 2x 1TB nvme 4x 22tb (104€ 0%VAT) SX135 2x 2tb nvme 8x 22tb (204€ 0%VAT) SX295 2x 8tb nvme 14x 22tb (~384 0%VAT)

    If you manually add the disks to configurations like AX-line, there is a limit, that may be bypassed if you contact hetzners support, but that'd be expensive compared to the SX line I reckon.

    Also, Hetzner's server auction has quite affordable 4x 16tb servers, starting from 63€ 0%VAT, that'd be pretty affordable RAID1- solution, ~1€ per TB/month.

    Unlimited traffic on all of the above, and a sidenote: SX line servers have one-off setup fees, server auction has no setup fees.

  • It's supposed to be a dot (.) character. The project's name is n.eko.

  • They're working on no-js support too, but this just had to be put out without it due to the amount of AI crawler bots causing denial of service to normal users.

    1. Doesn't run against Firefox only, it runs against whatever you configure it to. And also, from personal experience, I can tell you that majority of the AI crawlers have keyword "Mozilla" in the user agent.
    2. Yes, this isn't cloudflare, but I'm pretty sure that's on the Todo list. If not, make an issue to the project please.
    3. The computational requirements on the server side are a less than a fraction of the cost what the bots have to spend, literally. A non-issue. This tool is to combat the denial of service that these bots cause by accessing high cost services, such as git blame on gitlab. My phone can do 100k sha256 sums per second (with single thread), you can safely assume any server to outperform this arm chip, so you'd need so much resources to cause denial of service that you might as well overload the server with traffic instead of one sha256 calculation.


    And this isn't really comparable to Tor. This is a self hostable service to sit between your web server/cdn and service that is being attacked by mass crawling.

  • Yes, Anubis uses proof of work, like some cryptocurrencies do as well, to slow down/mitigate mass scale crawling by making them do expensive computation.

    https://lemmy.world/post/27101209 has a great article attached to it about this.

    --

    Edit: Just to be clear, this doesn't mine any cryptos, just uses same idea for slowing down the requests.

  • Selfhosted @lemmy.world

    Anubis - Weighs the soul of incoming HTTP requests using proof-of-work to stop AI crawlers

    github.com /TecharoHQ/anubis
  • And replaced the word "AI" with "Apple". ( ͡° ͜ʖ ͡°)

  • Umm, that is quite literally hallucinations what you are describing? Am I missing something here?

  • All models hallucinate, it's just how language models work.

    Do you have sources for this claim that Mistral's models are trying to deceive anyone?

  • In general, to everyone who finds Yacy as an interesting project, just give it a try!

    It's relatively light weight, and having millions of pages indexed does not take that much disk space, in my case: 3.5 million indexed pages is around 200 gigabytes only.

    Yacy is far from perfect, and it's an ancient project. But it's still alive and kicking strong!

  • Hi!

    I've been selfhosting Yacy for some years, even tho I rarely use it (I'm mostly using Kagi these days).

    But some tips:

    • Set up something like this to your browser, this sends Yacy to crawl pages that you visit https://github.com/JeremyRand/YaCyIndexerGreasemonkey .
    • Get familiar with blacklists and try to find some public ones to filter out bad sites and adult content.
    • Tinker with Ranking & heuristics -> Solr boosts to get results that fit your use case more.
    • And in general, tinker with all the settings you can find!

    And not directly Yacy related, but you can use your own Yacy through Searxng as well, even in 'private'(non P2P) mode.