wmf

15 ms·

451.
by wmf 4mo ago
Search has ads but Siri doesn't. And when Apple puts ads in Siri they won't be Google ads. I don't think Google benefits from this deal enough to be worth paying.
452.
by wmf 4mo ago
Tail risk. Only <1% of people get punished by their car's data. IMO that's still too much.
453.
by wmf 4mo ago
You might need different machines to cut wax/vinyl directly vs cutting lacquer to make a stamper.
454.
by wmf 4mo ago
This code creates a JSON intermediate representation that LLMs could probably consume. You might want to simplify it to focus on content and reduce token usage.
455.
by wmf 4mo ago
$4,000-5,000
456.
by wmf 4mo ago
Nvidia Spark has 10 mid cores and 10 fake cores. Probably all the others are faster.
457.
by wmf 4mo ago
UMA was never about performance and it still isn't. Spark is slower than a 5090.
458.
by wmf 4mo ago
This chip was designed before the shortages. I think they'll order just enough units to say they released it but not enough to put a dent in Rubin.
459.
by wmf 4mo ago
Nvidia has been making CPUs for 10-15 years.
460.
by wmf 4mo ago
Lemire is very narrowly interested in CPU SIMD so within that niche it may be interesting. As you said, overall the Spark is good but not great.
461.
by wmf 4mo ago
libgit2 exists. You could imagine communicating with some gitd over a pipe/socket but I don't know why that would be a good idea. Short of that you have to spawn processes.
462.
by wmf 4mo ago
The BDFL model is pretty well known and successful, especially for early-stage projects. I don't see any problem here. If anything, I think the public has been kind of brainwashed to expect multi-stakeholder style governance in spite o
463.
by wmf 4mo ago
At this price, GPUs might be more profitable than tokens.
464.
by wmf 4mo ago
With ~$2.1B/month of GPU rental revenue, is xAI now profitable? Are all divisions of SpaceX now profitable?
465.
by wmf 4mo ago
Based on previous patterns, the low should be over $75K. It's already broken.
466.
by wmf 4mo ago
In the real world layering isn't that strict. QUIC performs similar functions as TCP so it's also a transport protocol.
467.
by wmf 4mo ago
It's easier to undelete if you realize that the original takedown order was unjustified.
468.
by wmf 4mo ago
If you mean an OS kernel, it would be interesting to create a fully async kernel using async/await. I've only seen one project working on this.
469.
by wmf 4mo ago
Soft delete is the standard for many reasons and this is one of them.
470.
by wmf 4mo ago
I don't think loading up a single context window costs $1,500. Which limit are you talking about?
471.
by wmf 4mo ago
Discussion from April when this was announced: https://news.ycombinator.com/item?id=47760529
472.
by wmf 4mo ago
SDR is 24 bit and HDR is 30 bit. However there seems to be some brightness metadata required to specify how bright 1023 is. What baffles me is that there seem to be five or six different standards for this metadata.
473.
by wmf 4mo ago
At least there shouldn't be any complaints about benchmaxing this time.
474.
by wmf 5mo ago
Andy Jassy is actually trustworthy.
475.
by wmf 5mo ago
The new chip isn't out yet so that's the only thing they could be doing.
476.
by wmf 5mo ago
Oh boy, there's a whole report on this topic: https://www.cs.ucdavis.edu/~koehl/Teaching/ECS188/PDF_files/...
477.
by wmf 5mo ago
Also parallel construction. "We caught him due to reports from his bank." (That's not how they caught him but those reports did exist.)
478.
by wmf 5mo ago
It's fast enough if the bus is really wide.
479.
by wmf 5mo ago
Get ready for similar cases about the four-wheel Amazon delivery "cycles" in the US.
480.
by wmf 5mo ago
In other words, the Italian police are flooded with more reports than they could possibly investigate.

More ›