krackers

65 ms·

721.
by krackers 1y ago
It seems like is precisely what the other commenter implemented? https://news.ycombinator.com/item?id=44645423
722.
by krackers 1y ago
I had this exact question a few months back - https://news.ycombinator.com/item?id=44084781
723.
by krackers 1y ago
More than just the language, with objc on mac you also get to take advantage of the surrounding stdlib. Toll-free bridging between CF and Cocoa, libdispatch which provides intuitive primitives for concurrency, etc. It's a shame that de
724.
by krackers 1y ago
>you have to manually rebase every subsequent change What do you mean by this? You can do an interactive rebase in git as well. The real issue is non-trivial merge conflicts which is going to be an issue no matter you use.
725.
by krackers 1y ago
Has there been an "arms race" of sorts with programming contests, as knowledge percolates and harder categories of problems need to be chosen? Since these days dynamic programming is basically table stakes for ICPC-type problems.
726.
by krackers 1y ago
Also related https://epoch.ai/gradient-updates/how-has-deepseek-improved-... and some sections of https://semianalysis.com/2025/07/11/meta-superintelligence-l...
727.
by krackers 1y ago
Except generating more tokens also effectively extends the computational power beyond the depth of the circuit, which is why chain of thought works in the first place. Even sampling only dummy tokens that don't convey anything still pr
728.
by krackers 1y ago
That should be the tagline for their next tshirt~ https://swag.htmx.org/en-usd/collections/htmx-sucks
729.
by krackers 1y ago
This is about the visual design of "Kimi Researcher" (deepresearch equivalent from Kimi Moonshot) rather than the taste of their researchers
730.
by krackers 1y ago
At least for LLMs, the platonic representation hypothesis doesn't seem that surprising. The training data for models that are large enough to faithfully capture reality will necessarily have a lot of overlap. So in this weak form the p
731.
by krackers 1y ago
isn't this the page that used to have cursors everywhere in the background? I think the distracting design is some intentional running joke at this point
732.
by krackers 1y ago
>Which animal best represents your Firefox browsing style? Ironic that "fox" isn't even an option. And the fact that they even ask this tells that they probably don't want serious feedback.
733.
by krackers 1y ago
inb4 "We've simplified and streamlined the firefox experience by removing confusing control knobs and options."
734.
by krackers 1y ago
Assuming a 60Hz refresh rate, does it take ~16ms (± the vblank inteval) for a complete cycle from top-left to bottom-right? Or does the scan happen faster than that (with something else being the limiting factor on overall refresh rate)?
735.
by krackers 1y ago
Although thinking a bit more, even constrained to only output dots, there can still some amount of information passing between each token, namely in the hidden states. The attention block N layers deep will compute attention scores off of t
736.
by krackers 1y ago
Firefox should have appealed to the developer/hacker demographic. Instead they've made every effort to alienate those demographics in trying to appeal to the general public. They were never going to be able to compete against Goog
737.
by krackers 1y ago
From https://github.com/uBlockOrigin/uBOL-home/wiki/Frequently-as... >Keep in mind that uBO's own JavaScript-based network filtering engine has been measured to be faster than a well-known Rust-based
738.
by krackers 1y ago
The same way it's done with V3, because no permission-level blacklist/whitelist is going to prevent the person from creating an interpreter within JS itself. Looking at https://developer.chrome.com/docs/websto
739.
by krackers 1y ago
Fair, I was considering chrome headless since firefox users are already served google captchas more often.
740.
by krackers 1y ago
>an extension could spoof it not if they make it dynamic somehow (e.g. include current day in hash). Then with MV3 changes that prevent dynamic header manipulation there is no way for an extension to spoof it.
741.
by krackers 1y ago
One of the main goals of MV3 seems to be nullifying protection against tracking URLs. Most of the discussion about adblocking technically "still working" under MV3 misses this point. It doesn't matter if you're actually
742.
by krackers 1y ago
>They decided it wasn't a security issue, and honestly, I agree, because it didn't give extensions access to data they didn't already have. So they admit that MV3 isn't actually any more secure than MV2?
743.
by krackers 1y ago
In current chromium source, it seems still possible to force manifest v2 extensions with `kAllowLegacyMV2Extensions` feature flag? https://source.chromium.org/chromium/chromium/src/+/main:chr... This how
744.
by krackers 1y ago
Probably the results were worse than K2 model released today. No serious engineer would say it's for "safety" reasons given that ablation nullifies any safety post-training.
745.
by krackers 1y ago
Only HN would compare religion to L2 regularization! That said, your perspective "self replicating cultural program that benefits its biological host" is one that I think I've seen expressed before by the philosopher/cog
746.
by krackers 1y ago
interesting, I thought one of the whole points of o3 was mixed multimodal reasoning (e.g. everyone doing those geoguesser challenges). But maybe that's just a parlor trick and it's not actually implemented that way. I wonder when
747.
by krackers 1y ago
I thought so too back when the test was first released, but now that we have multimodal models which can take images directly as input, shouldn't this point be moot?
748.
by krackers 1y ago
Every time this discussion comes up, I'm reminded of this tongue-in-cheek paper. https://ai.vixra.org/pdf/2506.0065v1.pdf
749.
by krackers 1y ago
>what it is boosting Their profits
750.
by krackers 1y ago
There are several such papers, off the top of my head one is https://arxiv.org/abs/2404.15758 It's a bit more subtle though, if I understand correctly this only works for parallelizable problems. Which makes intui

More ›