Preferences

Reubend
Joined 1,716 karma

  1. Good catch. I didn't realize that there was a longer list of restrictions below the section called "Stricter mode", and it seems like a lot of String functions I use are missing too.
  2. When reading through the projects list of JS restrictions for "stricter" mode, I was expecting to see that it would limit many different JS concepts. But in fact none of the things which are impossible in this subset are things I would do in the course of normal programming anyway. I think all of the JS code I've written over the past few years would work out of the box here.
  3. Anything except a 3bit quant of GLM 4.6 will exceed those 128 GB of RAM you mentioned, so of course it's slow for you. If you want good speeds, you'll at least need to store the entire thing in memory.
  4. Maybe I'm totally misinterpreting, but the chart I'm looking at says "Net Win Rate of SAM Audio vs. SoTA Separation (text prompted)", so perhaps a lower number means that the alternative model is better?
  5. This provides plenty of value conceptually, but I wish they were able to push the syntax into a more intuitive place. My biggest gripe with Relay is how forced the syntax feels, and this seems better but still confusing. Take for example this component declaration:

    export const PostCard = ({ post: postRef }: { post: ViewRef<'Post'> }) => { ... }

    That feels terrible to me! I don't have any suggestion from my side, but I feel like there's got to be a less awkward way.

  6. It would be REALLY cool to see this same technique applied to a much more recent OSS model distillation. For example, Mistral 3 14B would be a great target. How efficient can we get inference there?
  7. Really impressive. I was surprised when listening to their demos how poorly their closed source competitors handled common abbreviations like "Wed. June 23rd". I had assumed that every commercial vendor had handled cases like that more elegantly, in line with how a voice actor would read them.
  8. I don't know what the book describes, but I'd like to hear more specifics. Until then, I'm going to assume that it's sensationalist hogwash.

    > The same knowledge that helps us treat neurological disorders could be used to disrupt cognition, induce compliance, or even in the future turn people into unwitting agents.

    Disrupting cognition is easy. But as far as I'm aware, we don't have any drugs to "induce compliance" and we're miles away from being able to turn people into "unwitting agents" purely on the basis of neuroscience.

  9. OpenAI likes to time their announcements alongside major competitor announcements to suck up some of the hype. (See for instance the announcement of GPT-4o a single day before Google's IO conference)

    They were probably sitting on this for a while. That makes me think this is a fairly incremental update for Codex.

  10. Before the outrage comments pour in, take a look at the conditions he has to meet (per the New York Times):

    > ...this 12-step package asks Mr. Musk, the company’s chief executive, to vastly expand Tesla’s stock market valuation — to $8.5 trillion from around $1.4 trillion — while hitting a variety of other goals. Those include selling one million robots with humanlike qualities and 10 million paid subscriptions to the company’s self-driving software.

    The headline $1 trillion figure only comes into play if he hits a very lofty goal: 6x the company's valuation and sell a bunch of expensive robots.

    I don't think this is nearly as crazy as people are making it seem. It's a huge reward for a goal that seems unlikely to be achievable in a short time frame.

  11. Despite having little understanding of the law in the U.K., I can confidently say that strangling someone without their consent want already illegal. There's no need to criminalize the consensual version of it.
  12. There are tons of musicians on Mac, and it gets lots of studio use. I'd say at least 50% of music studios are on Macs from what I've seen.
  13. Any company names or special acronyms should be added to your prompt.
  14. These are wonderful. They're so full of character. But I must imagine a "screening" on a TV would be a terrible experience to watch. I guess if you had never been to a full movie theater, you'd never know what you were missing out on.
  15. While we're all here, does anyone have some theories on simi8lar bets worth making? The article says that CZ's investments "practically screamed pay-for-pardon".
  16. Wouldn't you need a little material "left over" to claim that it can pass through itself? Two spheres of equal size wouldn't work because they would occupy exactly the same space.
  17. Can you give us some idea of the capabilities? What should we expect it to be able to handle? How long should we expect to wait for replies?
  18. Instead of all this stuff, I'd like to see Google use their ML chops to "solve" weather forecasting and deliver ultra accurate predictions a few days ahead.
  19. The tariffs have been a disaster for the American economy. What a tremendous unforced error.
  20. Ah, that's interesting. Are there any benchmarks of the Qualcomm chip versus the Vision Pro?
  21. I really wish they pushed for a 120 hz refresh rate instead of 90. IMO, this makes a huge difference for the immersion. I'm guessing that they didn't want to have stutters if their chip can't handle the higher FPS, but the refreshed Vision Pro will have a significant advantage there.
  22. Last time these folks were mentioned on HN, there was a lot of skepticism that this is really possible to do. The issue is cooling: in space, you can't rely on convection or conduction to do passive cooling, so you can only radiate away heat. However, the radiator would need to be several kilometers big to provide enough cooling, and obviously launching such a large object into space would therefore eat up any cost savings from the "free" solar power.

    More discussion: https://www.hackerneue.com/item?id=43977188

  23. interesting that their landing page doesn't use the word "browser"
  24. But non-motorists still benefit from roads.
  25. I'm not sure why the author thinks that no one's talking about these problems.

    I've heard a lot of people complaining about context loss from agents, whether that's due to context windows, or communication across agents, or agents not paying attention to what the user specified in their prompt in the first place.

  26. I really want to see how they figure out the actual delivery of the food.

    Obviously the cars can drive themselves on public streets, but how do you go up to someone's house and put a burger on their doorstep?

  27. What on Earth will you use as a justification for that? We pay taxes for roads, for fuel, and for the cars themselves. The last thing we need is more.
  28. > Zed does not implement Ctrl+Z yet

    Wait, what? That seems like an incredibly important missing piece of functionality.

  29. Looking through the meeting notes myself, I don't see MSFT mentioned specifically. That makes sense given their relationship to OpenAI as investors; they might be okay with OpenAI calling out "big tech" in general, but it would be extremely weird if they made any specific references given that Microsoft owns 49% of their company.*

    *technically it's some weird profit sharing thing rather than equity

  30. > The “realism” of graphics has nothing to do with performance.

    Obviously I understand your point that computational complexity is different than the extent to which something is realistic. But it's totally wrong that it "has nothing it do with" it.

    Photorealistic scenes require high res textures, higher detail levels in geometry, better shadows, better global illumination, etc...

    Cartoonish art styles don't necessarily require any of those. They still benefit from them, but with diminishing returns.

    It's cool if they want to take advantage of some fancy UE5 features, but the burden to optimize is on them, especially considering that the game's quality settings look like this: https://www.thegamer.com/borderlands-4-optimal-pc-settings/#...

This user hasn’t submitted anything.

Keyboard Shortcuts

Story Lists

j
Next story
k
Previous story
Shift+j
Last story
Shift+k
First story
o Enter
Go to story URL
c
Go to comments
u
Go to author

Navigation

Shift+t
Go to top stories
Shift+n
Go to new stories
Shift+b
Go to best stories
Shift+a
Go to Ask HN
Shift+s
Go to Show HN

Miscellaneous

?
Show this modal