@pervognsen There are only two modes in life, greek or assembly ;).

RT @dougallj: Who wouldn't want to read about the cursed ABI that lets you mix ARM64 code and X64 code in the same address space?. https://….

Generate videos in just a few seconds. Try Grok Imagine, free for a limited time.

If you're gonna shill F* just try to build it on 32 and 64 bit platforms.

Cattle, not pets, but also don't even think of breathing in prod's general direction.

I guess you never have to shell into prod hosts if you only ever debug known unknowns.

Nothing says pragmatic like writing a bunch of conditionals over 20 LOC for an operation that should be <5 instructions.

What's the ultrafinitist take on the "infinity outfits" question?.

Ah, rationalists: "it's the same game, they payouts are identical!". "After reducing two hypotheticals to 2x2 matrices with values in {-1, 0}, I don't see any difference between them. Debate me!".

But surely a ticket left open for more than 6 weeks will never be addressed.

Security researcher for a cloud provider that makes money by oversubscribing tenants thinks intra-core leaks are an important attack vector… to be fixed in microcode.

RT @ColumbiaCompSci: 🎉We are celebrating Mihalis Yannakakis’ contributions to science on the occasion of his 70th birthday with a 3-day wor….

"literate" python one-liner: `#!/usr/bin/env sed -re s|^|\x20\x20\x20\x20| -e s|^\x20{4}\x23\x23\x20?||`.

Passed Albany, started describing distances in hours again.

Hair straightener popcorn gal is from montreal. Barely surprised.

* bit 5 (wth brain?).

Fun interaction with SKX102 mitigations (.

Branch prediction tables on recent Intel are partitioned by jump address bit 31. Using that for isolation is such a good hack

We often have to strengthen the induction hypothesis in recursive algorithms… but it sometimes pays to weaken and restore what was lost in post. In this case: what if we wanted TopK in any order?.

(dans mon temps, sanction était féminin).

lecture de vacance:

I don't know about the takes on absurdly low p values < 1e-1000. ISTM the real interpretation is "the sampling isn't the dominant source of error, it's probably implementation errors or cosmic rays." And that's a useful result, because these error sources always lurk!.