Skip Navigation

Posts
1
Comments
177
Joined
5 yr. ago

  • The supply chain for computing resources is extremely bottlenecked right now. Even with the high demand for open weight AI models the data centers hosting them aren't able to get the computing resources they need and they keep running into rate limits even for paid users. Z.ai's hosting quality has dropped which I suspect may be related. Even over the past few weeks this has gotten much worse with the release of Kimi K2.5 being competitive with closed US-based models and OpenCode becoming popular. Meanwhile we have corporations like OpenAI buying up half the world's RAM fucking both other people and other corporations. So I'm not sure where this is going to end up, but the computer hardware market is going to really suck for a while.

  • A compiler doesn't have to make the exact same output, nor does an LLM have to make differing outputs, and a compiler uses someone else's code to transform your code into something else.

  • I'm okay with a few crawlers, but not what's effectively a DDoS attack by AI companies who abuse my resources generating terabytes of traffic and crashing my server while costing me money. I use Anubis now, which sucks from an accessibility standpoint but I'm not dealing with their malicious traffic anymore.

  • Yes if it's tunneled to your self-hosting setup. With CGNAT you have to use similar services if you want to self-host.

  • Stoat and Matrix do fine with that model.

    Having checked a year ago and checking again now, their UX has barely improved and the GUI is still significantly worse than Discord. They need some sort of monetization strategy for funding and leadership that isn't a shitshow so that they can make something that people will actually want to use.

  • https://reuse.software/faq/#uncopyrightable

    The REUSE specification recommends claiming copyright even if it's machine generated. Is this incorrect information?

    EDIT: Also, how is copyrighting code from an AI different than copyrighting an output from a compiler?

  • I like the bottom image

  • For coding $280/mo is peanuts compared to how much the Claude API costs

  • GO self-hosted, go local

    Last I checked it costed ~$6000 to run a high end LLM at terrible speeds and that was before the RAM price hike. And at the rate things are changing it might be obsolete in a few year. And I don't have that much money either.

    I'm going to stick with the free OpenCode Zen models for now and maybe switch to OpenCode Black or Synthetic or whatever when they stop being available and use the open weight models there for now.

  • Just log in with a third party client where people in the Issues are complaining about being banned. Or use your Discord token and spam their API with illegal calls until they ban your account

  • Matrix is sort of a mishmash of things, not a fan of the UX but it has federation and encryption and isn't Discord, and also has bridges to various platforms

  • That doesn't look like clarification to me. It explicitly mentions sexual content then gives a vague "age-restricted themes" and "18+ content" that they can apply arbitrarily like YouTube did with "Hind's Hall"

  • Youtube age restricts pro-Palestine stuff and the government age restricts the Epstein files, so I can imagine this would target political or LGBTQ spaces, which are some of the worst spaces to require ID for. There's a local political server I'm in, which due to this I'm strongly urging to migrate to Matrix with Discord bridges, and am offering to help with.

  • Reddit feels like Quora at this point with all the affiliate links and promotions

  • If I ever find myself there I'll just digg my way back to Lemmy

  • Now when people try to use Bing Copilot app to download a normal web browser, it can talk them out of it while introducing them to all the other Copilots and their apps

  • Ok then let's dissect your statement.

    Short term gain.

    Not a complete sentence so not even sure what this means. But when taking it into the context of my statement that you replied to (which was about the effectiveness of the protests), I assume you're saying that I was focusing on short-term gain in contrast to focusing on the long term when determining the effectiveness of the protests. Which hints that I should be taking different aspects into consideration and arguing differently. Which is a position, which I then assume you intend to defend.

    If the topic wasn't about the effectiveness, which was the primary topic of the comment you replied to, that should have been clarified.

    It’s more about long term awareness.

    Based on what's inferred from the previous statement, this further validates its intent. It also brings up what specifically should be considered when arguing differently, which is "awareness".

    It is unfortunate for those now. But

    "Some of what you said is correct, but..." Argumentative-style writing, reinforcing the intent stated earlier.

    it’s not something that changes overnight.

    Based on the context of the previous statements, reiterates that the effectiveness is determined by the long term, specifically "awareness", as opposed to how I measured it.

    So it seems like you were trying to justify something.

  • The way you initially commented made it seem like you were arguing that long-term awareness justifies its classification as "very" effective and I'm arguing that it doesn't. And now you're saying that you're arguing the same thing I am, so I can't even tell what your stance is at this point or why you brought it up.

  • But you crap on mine for understanding that I don’t have any better ideas too, but am aware of what protests do in terms of change? Got it.

    What's wrong with that? The title of this post was that protests are "very" effective. I'm arguing that they're not, and apparently so are you.