Hacker Newsnew | past | comments | ask | show | jobs | submit | more dwroberts's commentslogin

> Remember that uncanny valley between 1KB and 25MB? The problem isn't the sizes—it's that Second Age systems force you to choose between two bad options: inline the data (killing performance) or store pointers (breaking governance). Spiral eliminates this false choice. We store 10KB embeddings directly in Vortex for microsecond access, intelligently batch 10MB blocks of images for optimal S3 throughput, and externalize 4GB videos without copying a single byte. One system, no compromises.

No compromises but isn’t ‘externalising’ a large video the equivalent of storing a pointer in the first example? Can’t really see any other way to understand what that means (it goes to an external system and you store where it is)


The predictions don't tell us anything about why the answer is what it is. There is probably important (useful) fundamental scientific knowledge in being able to know that vs. just being able to predict the result.


You could always put the helm chart in a Kustomize and change the things you don’t like.

—-enable-helm isn't supported everywhere but Argo definitely allows it


We just went through this whole Kustomize shenanigan in our company. Seems completely asinine. Why not just fork the chart, fix it yourself?


... or the quite common case, make helm write the template once, fix, port to your own process, delete helm, live happy


> live happy

Until the next major upgrade.


In my experience, an update big enough to require major rewrite, probably should require a portion of this process to figure just what is the upgrade path.


> Note: VERBAL models are asked using the verbalized test prompt. VISION models are asked the test image instead without any text prompts.

Just glancing at the bar graphs, the vision models mostly suck across the board for each question. Whereas verbal ones do OK.

And today's example of clock faces (#17) does a good job of demonstrating why: because when a lot of the diagrams are explained verbally, it makes it significantly easier to solve.

Maybe it's just me, but #17 for example - it's not immediately obvious those are even supposed to represent clocks, and yet the verbal prompt turns each one into clock times for the model (e.g. 1:30) which feels like 50% of the problem being solved before the model does anything at all.


> Unlike conventional large language models prone to hallucination, these new systems leverage dynamic multi-method generation (DMG),

The only search results for this term are this article and promotional things for this company getting funded (Xavier)


Let me guess, it involves giving a bunch of different LLMs the same inputs, and picking the best output?


Probably they feed the outputs into another LLM.

Quick, patent it, save the prompt, that's golden IP right there.


The overuse of emojis in the comments makes this whole thing feel a bit off to me. Like someone faking interactions and trying too hard to make them look convincing


Forums with lots of women often use more emojis, its just how it is.


They address this and state that there is no reason to believe it wouldn’t also apply to the general population who do not have lipohypertrophy


However they have no evidence of this. While I would expect this to be the case (as a researcher of ageing epigenetics), this is entirely conjecture.


> It is rare to read something more moronic than that

It's not actually wrong though is it - real codebases have been implementing reflection and introspection through macro magic etc. for decades at this point.

I guess it's cool they want to fix it in the language, but as always, the approach is to make the language even more complex than it already is - e.g. two new operators (!) in the linked article


> been implementing reflection and introspection through macro magic etc. for decades at this point.

Having a flaky pile of junk as an alternative is never been an excuse to not fix the problem properly.

Every proper modern language (Rust, Kotlin, Zig, Swift, even freaking Golang) has a form of runtime reflection or static introspection.

Only C++ does not. It was done historically with a mess of macros or a pre-compiler (qt-moc) that all have an entire pile of issue.

> the approach is to make the language even more complex than it already is - e.g. two new operators

The problem of rampant complexity in C++ is not so much about the new features when they bring something and make sense.

It is about its inability to remove the old stuff even if it is consensual that it is garbage (e.g iostreams).


> Having a flaky pile of junk as an alternative is never been an excuse to not fix the problem properly.

Thank you. Some people use the phrases "real projects" and "production code" as if they imply some standard of high quality.


> but this did allow for a port of the game to the OG XBOX (733 MHz PentiumⅢ box) way back in 2003

Not sure if the clock speed is just for reference or emphasis re: efficiency, but RCT1 will in fact happily run on a Pentium 90 (which is still mind blowing to me given the scope of the game)


Just for disambiguation to emphasize that I'm talking about the Intel-based console, because the naming scheme of the later Microsoft consoles makes it easy to confuse “Xbox One” with the OG one. I spent most of my time playing RCT 1 and 2 on a 400 MHz PⅡ, and their performance was indeed flawless :)


Having cut my teeth writing asm on 386/486 in ms-dos, these comments are kind of hilarious to me because Pentium is well into "you can write most of it in C" territory.

By the P2 era (97-98), especially as consoles show up, assembly's not desirable at all.

Pmode/w was released in 97 which speaks to the demand for a Watcom C/C++ protected mode extender at the time...


I don’t think necessity has anything to do with it being written in assembly in the first place, it’s just Sawyer’s background was in porting others’ titles and it was just what he was used to using


> his lease said that he wasn't allowed to receive mail at the house he was legally renting.

Pretty sure that is not a stipulation you can legally put in a tenancy contract. Because both parties have to be able to serve notice on the other via post in writing. Same reason you are legally entitled to know the postal address of the landlord.


I'm sure you are right, but that didn't stop the landlord from trying their luck. Your observation about serving notice is on point, because in the end the deposit was returned only after my friend filed a small claims case against them.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: