HNHacker News
TopNewBestAskShowJobs

tliltocatl

1,072 karma · joined June 23, 2024

submissionscomments
tliltocatl··on EU ban on destruction of unsold clothes and shoes enters into application
At that point they might skip the bonfire part (or the locals would without asking their permission), which is kinda the whole point.
tliltocatl··on RISC-V Is Inevitable: State of the Union Keynote Argues
> the densest 64bit ISA

Because all other 64-bits were designed with no regard to density too. On 32-bit (and thats where density matters) it's sill worse then Thumb even with compression extension.

tliltocatl··on RISC-V Is Inevitable: State of the Union Keynote Argues
> I'm starting to get the feeling that there is something fundamentally broken in the RISC-V specification that fundamentally limits performance.

RISC-V is an objectively bad ISA design (a resell of MIPS dropping some of the most ominous features, then trying fix the code density issue with billions of extensions), but x86 is way worse and it didn't prevent Intel from making performant implementations. And RISC-V is certainly not bad in the way that would limit performance (well, maybe code density, but it's not the major issue).

The reality is, ISA matters very little for CPU performance. What really matters a lot is the memory subsystem and interconnect. I. e. good DRAM controllers IP are pricey - way more pricey than ARM cores AFAIK. Not a problem unique to RISC-V - i. e. Altera memory controllers used to be shit as well, not sure if Intel changed anything.

And another issue - there doesn't seem to be all that much money in CPUs any more. Look at ARM's history with high-performance sector.

tliltocatl··on Nvidia, CoreWeave, and Nebius: Inside the Circular Financing of the GPU Boom
By starving off access to cheap DRAM for everything that isn't AI?
tliltocatl··on eBay is down this looks like a global incident 07142026
MDY, seriously? Please don't.
tliltocatl··on Nvidia, CoreWeave, and Nebius: Inside the Circular Financing of the GPU Boom
> If market followed rational logic, you could pre-calculate which stocks will be what price and never lose money.

The fundamentals are unpredictable, so even a perfectly rational planning (suppose such thing could exist) would lose money sometimes. Not in the long run, but long run doesn't matter if a single wrecked ship can wreck you.

tliltocatl··on QuadRF can spot drones and see WiFi through my wall
I don't think it's any good for that. It's relatively narrowband and not the frequency you usually have issues with EMC on (5 to 6GHz - unless you are specially transmitting on this frequency you are unlikely to emit anything there).
tliltocatl··on Meta reuses old RAM in new servers with custom bridge chip
"Moore's law" is a marketing gimmick. The real physical law that held for decades is Dennard scaling, which stopped to apply already in 2006 once transistors got too small so short channel effects and gate leakage kicked in.
tliltocatl··on You Don't Own Your .io or .ai. You Rent a Country's Politics
Yea, but governance structure can be changed on a political whim. .io is efficiently governed as a gTLD now, even if it is a ccTLD. While .com/.net/.org are generic and governed by ICANN, the registry is VeriSign, a USA-based company, and ICE made it explicit already in 2010 they can seizure these any time they see fit. Back in 2010 it was mostly for genuine law-enforcement ends (whatever these ends justified the means is another story), but with what became of ICE recently, I would not bet on anything.
tliltocatl··on You Don't Own Your .io or .ai. You Rent a Country's Politics
Same applies for .com/.net, except it's USA's politics. Which you may or may not consider more reliable. But yea, ccTLD domain hacks are pretty damn stupid.
tliltocatl··on Do Wavy Walls Use Fewer Bricks? I Tested It in Blender
That's to be expected, in this test setup wall density is the only thing that matters (and Blender physics friction model is probably not quite realistic either). But heavy rolling ball impact isn't the kind load these walls are normally subject to.
tliltocatl··on North Korea dictator Kim Jong Un is illegitimate because his mother was Japanese
Except NK isn't a hereditary monarchy. One might say "so what, they still have a dictatorial rule by a single family", but it is an important difference: the leader post inheritance isn't governed by a clear set of rules so familial illegitimacy is just irrelevant for politics.
tliltocatl··on Anonymous GitHub account mass-dropping undisclosed 0-days
A friendly reminder that a 0-day is a vulnerability that wasn't known until after a malicious actor exploited it. If someone publishes a PoC, it is not a 0-day, just a vulnerability.
tliltocatl··on The unbearable cheapness of open weight models
Depends on the application, of course. For "google replacement" they are trying to sell it as - it's absolutely essential, but even then it works like crap. For coding… Maybe it's not so essential? Yea, it would not know about latest libraries, but maybe that's not much of a problem? Then, of course, it is a big question if LLM code generation is worth it at all.
tliltocatl··on Millimeter wave technology drills 100 meters into granite
Based on what I could find:

They generate the RF on the surface and transmit it down the borehole thru a waveguide, so it's only limited by arching in the waveguide. Since we only need power transfer and don't care about multimode propagation, the waveguide diameter isn't limited, and probably on the larger side to reduce copper losses. And the heat management is provided by blowing argon which also carries abalated rock particulate to the surface.

See the schematics here: the schematic here: https://www.thinkgeoenergy.com/wp-content/uploads/2021/03/mi...

tliltocatl··on Millimeter wave technology drills 100 meters into granite
Gyrotron isn't quite a maser, more akin to a free-electron (i.e. electron beam) RF source. AFAIR (might be wrong, but based on what I could find there: https://www.thinkgeoenergy.com/wp-content/uploads/2021/03/mi...) they aren't literally vaporizing the rock, rather locally heating it til it crushes into particles that can be blown away.
tliltocatl··on 8086 Segmented Memory was a good idea
Intel wanted to go full 32-bit with capability-like segments right after 8080 with i432. It took way too much time to design, so they released 8086 as a stair-step to 432. It still took too much time, so they released 286. And when i432 was ready it turned out painfully slow anyway and x86 was completely entrenched because of PC platform so they bite the bullet.
tliltocatl··on 8086 Segmented Memory was a good idea
It might have worked better if x86 had general-purpose registers where every register could work as a segment. Or maybe just many more segment registers. But with only two data segment registers to play with and quite cubersome (and slow!) loads, most software just chose not to bother.
tliltocatl··on Has AI already killed self-help nonfiction books?
"Self-help nonfiction" have always been a waste of paper. And honestly, most of the time I hear "X was replaced by AI" I find myself thinking "good riddance, but we could drop X altogether and not loose anything of value."

Fiction, on the other hand… Much of fiction's value isn't just the content itself, is that they create a shared language medium. A book might actually be meh (came up with some examples, but then decided to drop it so as not to offend anyone), but the fact that people you talk with have read the same book and understand same references makes reading it valuable. So, it's unlikely to happen, until we delegate all of our communication to AIs, which isn't likely to happen any time soon.

tliltocatl··on Making glass-to-metal seals for home­made vacuum tubes
Just keep it liquid, that's the point. If it freezes, it bursts the glass, as the OOP says.
tliltocatl··on A backdoor in a LinkedIn job offer
I wonder if a cease and desist to their legal department would work better?
tliltocatl··on Making glass-to-metal seals for home­made vacuum tubes
To everyone screaming "this would not work": it would work (with right kind of fluorine rubber) as long as the lamp is attached to the pump. Thermionics (but not CRTs) don't actually need UHV, HV is good enough, so no copper conflat necessary. If you want to seal the tube off - then yes, the tube needs a proper seal and a getter. Afaik, the getter is there to counter slow outgassing of tube internals (virtual leaks) rather than any external leaks.
tliltocatl··on Making glass-to-metal seals for home­made vacuum tubes
Than you can skip the hollow state part altogether. And any plastic-package parts would screw your vacuum up badly - if it survives sealing and bakeout that is.
tliltocatl··on Making glass-to-metal seals for home­made vacuum tubes
One thing about gallium/galinstan - it would actually make a descent high vacuum seal as it has lowest vapor pressure of all elements - so it doesn't evaporate. The problem is that it sticks to just about everything that isn't PE/PTFE. Galinstan thermometers use some proprietary coating to make glass repel it.

I was once entertaining the idea of using gallium for an electrostatically or MHD boosted Sprengel pump, but figured out sticking would make it infeasible. And now it's unobitanium too.

tliltocatl··on Making glass-to-metal seals for home­made vacuum tubes
Incandescent bulbs used dumet/platinite, which is an nickel-iron alloy like kovar except it's turned for a different CTE. This stuff isn't that expensive when mass produced - it's just those who can afford usually borosilicate can afford paying a premium for kovar.
tliltocatl··on Making glass-to-metal seals for home­made vacuum tubes
If you want a gas discharge tube and not vacuum, you can even drop the coil on the inside: https://hackaday.io/project/194683-plasma-toroid-sky-guided-...

But most hollow-state devices run on either DC or pulses, so coupled inductors wouldn't work.

tliltocatl··on Yon – a topos-oriented language with a content-addressed lattice heap
I really appreciate this comment. It is easy to dismiss something as AI slop when it looks and smells like AI slop, but pinpointing actual errors is much more valuable. As solomonb said in a sibling comment, "Imagine someone honestly interested in learning about category theory".
tliltocatl··on Moving beyond fork() + exec()
The problem with POSIX IPC is that passing file descriptors between processes (other than parent passing to child via fork) is hard. Yes, SCM_RIGHTS can do it, but it is quite error prone and rarely done.
tliltocatl··on Moving beyond fork() + exec()
In addition to what you said: forking from a process running on multiple cores is slow once you have mark all pages as read-only and shoot this out to all cores. TLB synchronization is super expensive. Unix originally didn't support threads (want concurrency? just fork!) but with modern multicore that's clearly unsustainable.
tliltocatl··on Ask HN: What was your "oh shit" moment with GenAI?
Still haven't had one. It is impressive, it is sometimes useful, it will be insightful (once the smoke settles), it is nowhere close to become self-improving world-as-we-know-ending ultimate solution to every problem it is being sold as. And much of the progress we have seen so far relied on tons of natural data being available thru the Web. After LLM killed SO, where would we get the answers to train LLMs on?
← PreviousPage 3 of 16Next →