NkVczPkybiXICG

I’m not informed enough to have an opinion on prop 40, but it’s disingenuous to say that she would have to sell $250m of her stake in in-and-out. She could take out a loan against her stake and pay the tax with the proceeds, then pay off the loan over time with the income from her ownership of in-and-out.

Big tech companies don’t usually pay hourly.

Why does the title say 1ns but the body of the article says 30ns (with hopes to eventually get it down to 5)?

Hint for more wins if you want it: which numa node is the vvar page on? Which numa node are you on?

In what way, exactly, was the gzip compression “clean”?

I was confused by this because pi is not between 1 and sqrt(1/2). But I found https://www.cs.miami.edu/home/burt/manuscripts/gaussagm/agma... and it clarifies that this AGM is used in the course of approximating pi not that it’s sufficient on its own to do so.

55kph is about how fast Remco Evenepoel just did a flat 12km time trial in the UAE tour, to win it. I think you’re being a touch overconfident if you think you can hold that for an hour.

You’re aren’t x, you’re just y llm slop

I think they mean TFLOPS, not FLOPS

Have they tried paying more? I know lots of people that would rather make chips than software but software pays better.

FTX didn’t make any of the architectural decisions presented in this talk.

Been great for me. Gets out of my way and I mostly don’t have to think about it. I’ve gone months without having an affirmative thought about it one way or another.

Two words: spider rain

https://www.nationalgeographic.com/animals/article/150518-sp...

It uses select, which is notorious for poor performance (although is relatively portable). Try not to use this project for anything high-load.

...or just move to Madagascar.

How about 500 USD to be settled in bitcoin (or similar virtual currency)? The metric for winning will be where the top (by market cap) 10 finance companies are headquartered at midnight Apr 28, 2034. If more are in NYC, I win. If more are in San Francisco, you win. If it's a tie, the "next highest" will be added until it isn't a tie.

Deal?

When the orders were added, there was a modification to the FIX specification for the exchange. For example: http://www.directedge.com/Portals/0/docs/Direct%20Edge%20Nex.... It is everyone who trades on that exchange's responsibility to stay up to date with what's available, and how it interacts with their trading strategy.

"concealed", in this context, just means "too lazy to carefully read the documentation".

Sorry. I said some pretty silly things.

See the new version of the posts. Hopefully it's a bit more coherent now.

Everyone on that exchange has the same access to the same types of orders. HFTs only get priority if someone gives them priority. If these order types are so powerful, why aren't everyone using them?

Also, I will reiterate: HFTs NEVER get fills on quotes that are worse than any other participant on the same exchange. That violates RegNMS, and is straight up illegal.

Sorry, that was vague. My point was that market participants who make many trades will rapidly converge on their mean return, with little fluctuation. If you sample a normal distribution 10M times with a mean of 0.01%, you will very rarely see a negative sum of the samples.

Undocumented? http://cdn.batstrading.com/resources/membership/BATS_FIX_Spe...

The "public price that everyone else sees" is the same price as the HFTs see. These are available on the direct feeds from the various exchanges, and there's no discrimination against non-HFTs. Anyone who pays for it can get it. It's an equivalent advantage to having a Bloomberg - more data, faster.

The quotes from HFTs are NOT inferior to existing quotes. At worse, they are identical to the best quotes already available on that exchange.

Law of large numbers. This is not surprising.

EDIT: Law of large numbers. Not central limit theorem. I blame alcohol.

There is no front-running: http://clarkgaebel.com/you-are-not-being-front-run/

TNs aren't for internships. J1s are.

He's the creator of Javascript, so obviously it is.

POWER7 has a max clock of 4.25 GHz according to that reference.

Anyhow, clock rate in this case is irrelevant. Look at the instruction latencies. I don't have these handy, but I'd bet $50 at at the chance of winning $10 that decimal floating point instructions (at least the divide) are slower than the IEEE754 ones.

These chips are also ridiculously expensive, so probably not the best benchmark.

Casting back and forth is extremely slow. He has a loop with multiplication by 10 inside it in dec_new.

Because it's easy to do things in base 2 with binary signals (like the kind used in computers). It's unnatural to use base 10 for anything in this domain.

It would be extremely unintuitive if math could be done faster in any base except for 2 (or a power thereof) when running on a modern CPU.