Viewing profile — adgjlsfhk1

right, but that's a reticle size chip. to put something in a phone it has to be ~10-30x smaller

The size of model we're talking about running doesn't need much if any dram.

I don't think this works out from a cost/silicon perspective. Small models already run pretty well in software (since the weights fit in cache) and big models require silicon area …

Heat is really useful if the energy form you want is heat. If you want electricity or motion it's not the best.

the GPU competition is using 16 gpus, so the actual comparison is that the CPU has <1/10th the bandwidth

> Federal prosecutor success rate is over > 90% Prosecutors pick their cases. Defense doesn't. The cases that aren't 90%+ sure aren't charged.

I think HDR wasn't what changed. It's resolution. At 720P and below, you don't notice dither because pixelation is the more obvious problem. Once you're at 1080p (and it's more obv…

The counterpoint is that you should just keep your images at 10 bit minimum or 12+ bit until the OS/monitor can dither for you. Lossy compression will do much better with a 10/12 b…

red/green/blue

Max effort is the way to give the highest perf, but not highest perf/$. Having claude (or other models) use a lower effort can often be 80% as smart but get to the results 10x fast…

Testing at max effort likely doesn't produce optimal results.

right, the thread you responded to is asking why there hasn't been a parallel criminal suit

that definition only converges for Re(z)>1. for Re(z)<=1, you need alternate formulas

Julia has roughly 2 different IRs that live pre LLVM (one for type inference, one for Julia level optimization). This is a proposed replacement to both of these. The reason not to …

median of median does give you guarenteed n*log(n) but it doubles your memory reads per pass making it pretty poor. single random is almost guarenteed to take fewer passes (and med…

it's arguably more of a cpu bug than a computer bug. The problem is that predictable data determined whether a cmov or a branch is faster. cmov is only faster than if when the bran…

if you have decent (randomized) pivoting, you never hit the worst case or anything like it

the problem with oral societies isn't that they unable to transmit information (in the Shannon sense) it's that they're really bad at transmitting information through time while se…

Binface's speculation to the media is that the Monster Raving Loony party may split the vote with Farage instead as that is closer to where his loyalties lie.

> SpaceX started moving into their own constellations because lower prices open up new markets, and this lets them access those markets without having to lower launch prices the al…

it's not just that they have the same asymptomatic. once you do the radix conversion (e.g. base 10 has log(10)/log(2) times more symbols), any 2 definitions are only off from each …

none of that explains why the dense states (e.g. ma, ny, ca, tx) have this solved

if the proposal was ml-kem+ecc hybrid that removed sha2 and replaced it with sha3, no one would be objecting

if no one should implement it, why standardize it?

yeah for real time signals higher frequency makes sense (very briefly before you fft and kill the high frequencies), but for stored signals nyquist is king.