xlrndo

Why do we think New Glenn had the spare performance to operate with an upper stage engine out? The reduced thrust doesn't just make the burn take longer; it takes more propellant, also. Unless the BB7 was well under the actual, real-world capacity of NG, gimbal limits or guidance are irrelevant.

Oh and Also I should say, it's entirely possible that GS2 doesn't have the gimbal range to support an engine out scenario. The old dual engine centaur didn't have engine out capability, the engines couldn't gimbal far enough to place the thrust vector through the center of mass if an engine failed: thespacereview.com/article/1321/1 However, I'd assume, perhaps naively, that modern GNC would make single engine operation much more viable so there'd be incentive to design the stage to support this.

Question for those more expert on rocket engines than me: if the BE-3U fires at low thrust, does the stage try to fire longer to compensate? How plausible that it fired the normal amount and had prop left over for a preprogrammed deorbit burn, albeit now in the wrong place?

@vashikoo That printer makes the same sound as a Star Trek communicator, and the rotating bed sounds like a Star Trek door.

This two-hour 1968 made-for-TV motion picture was the debut of a new TV series that ran for 12 years. Here's the wild intro scene and credit sequence for that famous show...🌊

You can store an index to pick which of the top-K tokens is output from the LLM. If it's very good, the index will almost always be 0, and only occasionally be further down. Compress the indices with arithmetic compression and the common case will only use a fraction of a bit. In very rare cases where the LLM doesn't have the right token at all, you can just store it.

It is generally frowned upon to have LLMs precisely regurgitate part of their training set, but it is an interesting question how you could use LLM training to nearly losslesly compress a huge corpus like the entirety of the Internet Archive. The Hutter Prize is for perfect compression, but only one GB. There would be different trades at the PB level, and it gets much more interesting when it doesn’t have to be bit-accurate.

@CSI_Starbase @EricTheEpic0403 @HalcyonHypnotic @rocketrepreneur Wouldn't have to add heat (at least no more than bubbling would add via viscous dissipation). They could have a passive spinner, magnetically driven from the outside. Any heat from electrical losses would be outside the tank.

@xlrndo @EricTheEpic0403 @HalcyonHypnotic @rocketrepreneur Mechanical components add heat. They want to remove it.

I wonder if this ends up impacting rocket launches. While some cryogenic stages are working toward autogenous pressurization, my understanding is that most of them currently use helium in COPVs for pressurization...

@CSI_Starbase @EricTheEpic0403 @HalcyonHypnotic @rocketrepreneur Why not use some form of mechanical agitation (impellers, etc.)? There are definite complications involved like material compatibility, how to deal with bearings, etc., but it seems like a solvable problem.

@EricTheEpic0403 @HalcyonHypnotic @rocketrepreneur It’s not subcooling where this is used, it’s for destratification And yes it’s vented, I spent a full 2 weeks looking for a recovery system but they don’t have one. I think it’s just incredibly difficult to capture and no matter what you store it in, it leaks constantly.

@MuseumCommodore The cool kids had an Epyx 500XJ joystick. We were still confused by all the pirated games we didn't have a manual for, but we were confused in greater comfort.

In the 80s there were SO many Commodore 64 games that half the time I had no idea what the hell I was supposed to be doing. Ocean’s 1985 The Transformers was the perfect example. I slapped in my copied tape (yeah… we all did it), waited through that beautiful loading screen, and suddenly I’m Optimus Prime… or maybe a random Autobot… or just a very lost truck? No manual. No instructions. Just pure chaos. I spent most of the game transforming back and forth like a confused robot having an identity crisis, while Decepticons hunted me down. Was I meant to shoot? Ram them? Or just look cool in vehicle mode? Who knows! The manual was still sitting in my mate’s bedroom in the next town. Why were so many Commodore 64 owners constantly in this dilemma? Because we were all running pirated copies. No manuals, no maps, no clue. Just a dodgy tape, a Quickshot or Competition Pro, maybe a TAC-2 joystick, and the hope that smashing fire and yanking the joystick in every direction would eventually work. More than meets the eye? Mate… I barely lasted 5 minutes.

@Noahpinion Just switch to M14+. Then the string only grows logarithmically with the number of minority groups. If even that's too bad, we can switch to Knuth up-arrow notation.

The real AI pause will be increasing the length of these abbreviations until AI has to spend $1 billion in compute costs just to say them each time

I've never understood these kinds of demands where something was allegedly promised and not delivered. If the tools don't work for you, don't use them (but consider that they might work a few months from now). I don't spend time thinking about whether some "promise" (which is usually just a vague prediction) did or didn't pan out.

@xlrndo @xwanyex I was promised AGI. I can't give the LLM an arbitrary task and expect it to complete it reasonably. Even of the small subset of tasks it can complete, it needs significant assistance and supervision.

One thing people say to AI skeptics a lot is, “that was true six months ago, but it’s no longer true” but if you’re saying that of the same claim over and over again for years, then somebody is wrong somewhere. It can’t always have been six months ago in reference to the same claim.

The problem is basically the same as the old quote "It is difficult to get a man to understand something, when his salary depends on his not understanding it." Anyone that's thinking clearly can understand that LLMs can fix real bugs. But a lot of people are engaging in highly motivated reasoning that prevents them from thinking clearly. And the people closest to the problem are the most motivated of all.

Again, this is just a case of there being nothing wrong with the argument, but something still being subtly wrong with the way it’s presented, something off. I don’t understand how there could be a software engineer alive who didn’t already believe that there existed bugs in very old, well-tested software that had never previously been discovered. And I fail to understand how anybody would have any intuition other than that advanced LLMs would be able to uncover such bugs. I feel like you could put two and two together on this the first time you ever watched an LLM spit out code.

@oldyzach I could hear the music before I un-muted the video. As an anti-piracy measure, the game asks you various words from the manual when you start it. I used to have every one memorized.

That was a very good selection of vehicles, right? 🔊 Stunts a.k.a. 4D Sports: Driving (1990), MSDOS

@romanhelmetguy They sniffed the glue instead of using it to bond them together.

It’s perfect.

@avidseries 8: Eliminate advanced courses completely so that students cannot get too far ahead (see Algebra I in San Francisco middle schools)

7 common "antiracist" policies in education: Lower graduation standards so more blacks graduate Reduce academic rigor so fewer blacks drop out Relax grading standards to try to reduce race GPA gaps Decrease SAT g-loading (make the test dumber) to attempt to narrow race score gaps Eliminate SAT and GRE requirements to hide the evidence showing the institutional and systemic anti-white/Asian racism in college admissions Admit black students to college and graduate schools with far lower academic qualifications than whites and Asians Dumb down, lower the minimum pass thresholds, or eliminate professional board exams

@benryanwriter I think his point wasn't that nothing happened in that era, but that the time seems so vague, and the history is poorly taught. How many people even know of the Spanish Flu? WW1 is about the only well-known event of the time but even then it gets about 1% the attention of WW2.

@xlrndo Ah, who knew.

Andreessen is woefully ignorant of The Age of Innocence, Anne of Green Gables, Laura Ingalls Wilder, Lonesome Dove, and The Gilded Age.

@benryanwriter The Oregon Trail game was set in 1848.

It’s as if no child ever died of dysentery on the Oregon Trail.

@NickChapmn @Jonathan_Blow @_trish_xD Yes, but when I enable memory leak detection I want every address returned from malloc to be freed exactly once. I also want malloc success to be strictly based on null or non-null. I shouldn't have to say that null is ok only if the size was 0.

@xlrndo @Jonathan_Blow @_trish_xD free can be called on null though

malloc(0) is legal C. let that sink in for a second. some compilers return NULL. some return a valid pointer you can't dereference but CAN free(). both behaviors are correct according to the C standard. you can allocate zero bytes of memory, get a pointer to nothing, and then dutifully free that nothing. and the language just shrugs and says "yeah that's fine." this is why C developers have trust issues.

@Jonathan_Blow @_trish_xD No argument there. Though given that it is implementation-defined, one would think they could just pick the better behavior in one of their language standard updates. It wouldn't break any existing compliant programs.

@xlrndo @_trish_xD I agree (in my language it is strongly-defined) but you have to remember that in the early days of C, there were all kinds of different computers and we hadn't settled on how to do a lot of things, yet.

@Jonathan_Blow @_trish_xD It should not have been implementation-defined behavior, though. It should have been guaranteed (assuming allocation success) to return a non-null pointer which could then be passed to free.

No, this is the most correct and best behavior because it does not produce control flow singularities. Would you rather need to have an if-statement every time you allocate a dynamic number of bytes? Please do not tweet stuff like this without a lot of experience systems programming….

@Scobleizer It spooked my cat first, who ran under the bed before I noticed anything amiss. Then the phone started buzzing like mad. Then I felt the shake.

We had a good shake a few minutes ago. Didn't even wake the kids up. I didn't even feel it. But Maryam said the bed was shaking. I was in the shower, hate when the bed is shaking without me in it. :-)

@DJSnM Teledyne is one of those companies that, like Ball, is somehow better known for a completely unrelated product. In their case, shower heads and dental hygiene devices.

I’m sure lots of engineering and testing went into this structure