Me1000

It’s a PR on their private fork, they’re not expecting to have this accepted upstream.

Stopped using it after about a week or two of usage. The only interesting use case was screen mirroring from my Mac, but that wasn't compelling enough to endure the weight of it on my face. I expected watching a movie would be a good use case, but in reality the brightness of the screen would reflect (I guess?) off my face and create a glare... so it ultimately wasn't a good movie device. Gave it to a friend who was excited about it, and he also stopped using it after about a week or two.

A native application that further locks users into some single platform? Or accept all the maintenance and development costs and burdens that keep the application one step behind Photoshop if they wanted to support multiple platforms?

Cursor's statement on the deal (which does not mention the option at all): https://cursor.com/blog/spacex-model-training

Wouldn’t it be better to use a grammar in the token sampler? Tuning is fine, but doesn’t guarantee a syntactical correct structured output. But if the sampler is grammar aware it could.

You can send messages to null, sendings messages to a deallocated pointer is going to be a bad time.

Then you go to another supplier. But any company with proper counsel will tell them the same thing: don't break the law, which is exactly what they're trying to coerce Anthropic into doing. DoD requests do not supersede the law.

ADP is reporting only 22,000: https://adpemploymentreport.com

You should read the book! :)

The premise of this study is a bit misguided, imho. I have absolutely no idea how many people _post_ harmful content. But we have a lot of data that suggests a _lot_ of people consume harmful content.

Most users don't post much of anything at all on most social media platforms.

Maybe it's a lack of imagination on my part, but how do spammers abuse self-hosted runners?

Hi! Congratulations on the launch. Is your intention to ship using WebKit on Window and Linux too?

Claude has a sycophancy problem too. I actually ended up canceling my subscription because I got sick of being "absolutely right" about everything.

That link says "Content Unavailable" for me, possibly region locked? I changed the "mx" in your link to "us" but same error.

This is an old relatively low budget TV movie, it's not on TV+ (which I subscribe to). Nor can you rent it on iTunes, it doesn't even show up when you search for it. Same for Prime Video, etc.

Not OP, but yes believe it or not it's impossible to find certain movies anywhere other than pirating them. One example is "Pirates of Silicon Valley", I watched it when I was young and recently wanted to watch it again. I pay for basically all the streaming services, I'm would have been happy to rent it from any service at all. I spent several hours trying to find a way to pay to watch it and never could.

I'm pretty sure it's just being used as a turn of phrase here. Under the bridge means commuters are riding BART, over the bridge means they're driving.

There’s an important distinction between the open weight model itself and the deepseek app. The hosted model has a filter, the open weight does not.

Yet it's interesting how we put the blame and punishment on the people being taken advantage of, and not the employers who are exploiting them. If both parties are breaking the law shouldn't we at the very least ensure that the business owner who is exploiting any number of workers is held to the same standard as an undocumented person whose only crime was not having the proper paperwork?

This is a technology demo, not a model you'd want to use. Because Bitnet models are only average 1.58 bits per weight you'd expect to need the model to be much larger than your fp8/fp16 counterparts in terms of parameter count. Plus this is only a 2 billion parameter model in the first place, even fp16 2B parameter models generally perform pretty poorly.

I'm also confused by that, but it could just be the model being agreeable. I've seen multiple examples posted online though where it's fairly clear that the COT output is not included in subsequent turns. I don't believe Anthropic is public about it (could be wrong), but I know that the Qwen team specifically recommend against including COT tokensfrom previous inferences.

This administration under through Elon is pushing to cut 50% of NASA's science funding. Mapping galaxies we'll never visit is a purely scientific endeavor. Trump seems to care more about military expansion or for lack of a better term more "masculine" expansion of space. The science stuff is not interesting to him, and I'm honestly not sure I think Musk cares about it that much anymore either.

I'm not a hater (or OP), but I think it's because GLP-1s are often talked about as if they're a kind of miracle drug. And it's very rare that drugs don't have some kind of side effects, especially after long term use. They might not even be universal, and we might not know what they are for years.

Yes, I understand there's an obesity epidemic, I also understand that GLP-1 drugs can have benefits outside of overeating. But with any drug, it's worth being thoughtful about its use.

Although I haven’t used these new models. The censorship you describe hasn’t historically been baked into the models as far as I’ve seen. It exists solely as a filter on the hosted version. IOW it’s doing exactly what Gemini does when you ask it an election related question: it just refuses to send it to the model and gives you back a canned response.

You’re just seeing a short summary of it, not the actual monologue.

Didn’t go there but worked with a lot of great folks who did. The main thing I think is that they require their students to get industry experience before they graduate.

I don’t remember the actual requirement, unfortunately.

The 32B parameter model size seems like the sweet spot right now, imho. It's large enough to be very useful (Qwen 2.5 32B and the Coder variant our outstanding models), and they run on consumer hardware much more easily than the 70B models.

I hope Llama 4 reintroduces that mid sized model size.

It's not like ChatGPT (or ChatGTP as half of people call it) is much better.

This is the real value here. Keeping a secure environment to run untrusted code along side user data is a real liability for them. It's not their core competency either, so they can just lean on browser sandboxing and not worry about it.

They explain how it's calculated, you just have to trust their calculations are correct.