I’m still glad we have Wikipedia as a foundation instead of a publicly traded corporation. The fact that there are way worse offenders doesn’t exempt them from scrutiny, though.

I wouldn’t trust them enough to run anything important.

Qwen 3.6 still works fine on my gaming PC that didn’t cost anywhere near that. ¯\(ツ)/¯

Open weight is analogous to a compiled binary. Similar to how Windows is closed source, but still runs in your own hardware, whereas Linux is truly open source.

So I’m not sure what the marginal cost per launch is. I remember reading that the propellant between Starship and Supeheavy is something like $100M per launch, which would put 20 launches at around $2B and would make the re-usability not actually a cost saver vs. an expendable rocket. However, I just read the article below that says the launch cost is more like $3-$5M, so I may have been completely off base in my original assumption.

https://spaceorbitals.com/articles/starship-v3-cost-per-kg-2026/

deleted by creator

“OpenAI” signed the open weight pledge, which only makes sense given the company name and original values. They’ve had gpt-oss for quite a while, but it feels like a pretty half-assed effort for the purposes of virtue signaling. I’m not holding my breath for them to release an open weight model on par with GLM, Kimi, or Qwen.

Yeah, there’s a lot of innovation going on to run powerful models on modest hardware and the state of the art for running local models is changing all the time. If the trend continues, models hosted in a data center will only be for niche use cases.

I thought SpaceX was at least on the right track with Starship, but then I read this article talking about how even one trip to the Moon could take up to 20 launches:

https://www.nextbigfuture.com/2023/12/nasa-says-up-to-20-spacex-starship-refueling-launches-per-moon-mission.html

deleted by creator

I’m also surprised they compare themselves to Steam/Epic when DRM-free is GoGs whole thing

I get the feeling that a lot of the motivation for developing something like this is to play “DRM-free” games that weren’t originally that way. 😉

What a shame it would be if the trend continued towards powerful free and local models that can run on hardware people already own. Then all the poor companies trying to crash the economy and rent their models to everyone at exorbitant prices along with all the data centers being built might go bankrupt.

Good point, that should theoretically be double the decode speed, though not sure how the prefill would differ since it’s mainly compute-bound.

I’ve had this happen when plugging the laptop into HDMI and not switching back to speakers before unplugging. Plugging back in the HDMI and switching to speakers has fixed it for me in the past. I’ve also had a cold boot fix my audio in the past.

Seems like an awful lot of trouble to save $100 not buying a 5060 Ti that also has 16GB.

I don’t generally take issue LLMs used as a tool, but I do have a huge problem with lazy slop slingers. I also don’t like that the frontier models are closed source and rent-seeking, especially when they were trained on copyleft code and by all rights should themselves be open source if they were respecting the licenses. I’d think Linus would have something to say about that.

IMO, a decent philosophy is that LLMs can be useful tools, but if anyone can tell you used a LLM, you failed. People should have a healthy fear of being ridiculed for outsourcing their critical thinking. Anyone using a ton of tokens shouldn’t be commended; instead, the quality of their work should be called into question because they’re likely using LLMs as a crutch instead of a tool. Commits by Claude probably mean the person didn’t review the diff to clean up the slop, and also probably didn’t understand the changes well enough to write a clear commit message themselves. I wouldn’t want to see any commits crediting Claude in the Linux kernel.

Edit:

I’d also think that if Linus likes LLMs and wants to see quality tech democratized, he’d be advocating for FOSS LLMs under the Linux foundation where all code and training data are open source. What we have now with Anthropic and “OpenAI” is the equivalent of cloud-hosted Windows, and open weight models like Qwen are more like Windows XP in that they can be run locally, but they’re still proprietary and can’t be inspected, modified, or built upon. We need the Linux of LLMs.

Oh, nice! I haven’t been able to find one because searching for “open source llm”, etc., always yields open weight LLMs instead of open source ones.

At least the models like Qwen have open weight versions. It’s the same concept as distributing compiled binaries, though, whereas we really need true FOSS models where all of the code and training data are available under a permissive license. All of these models were trained on copyleft-licensed content, so all of it should be FOSS if the licenses were actually being respected. From that perspective, distillation attacks shouldn’t even be necessary and I couldn’t give two shits that there is no honor among thieves when the real thievery is that these models are closed source.

If it’s a subscription that should really be a one-time purchase, best to look for alternatives. Especially with software subscriptions, there’s often an option that is completely free that does the job well enough, if not better.

I’d been using Linux at work for many years, then self-hosting on Linux. I wasn’t too keen on desktop Linux because popular distros default to Gnome, which I thought was pretty meh, and it wasn’t until I tried Mint that I was compelled to make it a daily driver. I’ve distro hopped a lot and always come back to Mint.