orbital-decay

10 ms·

91.
by orbital-decay 2mo ago
The thought of the fishermen watching tiktok on how to poach is ludicrous. They're perfectly aware about this method and its high illegality. It might be new to some HN readers who aren't going to do this anyway, though. >how
92.
by orbital-decay 2mo ago
https://www.unslop.news/ https://hcker.news/?ai=exclude
93.
by orbital-decay 2mo ago
Blast fishing has been practiced since at least early 20th century. Essentially everyone knows about it and understands how destructive it is, no social media required. Some people just don't care.
94.
by orbital-decay 2mo ago
What would you consider as good evidence of a backdoor? Apple got plenty of media coverup for their part in Operation Triangulation, and even hardware experts were defending it as a fuckup rather than a backdoor, providing it with great pla
95.
by orbital-decay 2mo ago
BBSes were the first, of course. In particular, Libgen, Sci-Hub and others can be traced back through several generations of libraries to the SU.BOOKS FidoNet echo conference created in the early 90's.
96.
by orbital-decay 2mo ago
KL is a credible shop, they basically founded the modern anti-malware industry and pioneered most basic techniques in the 90's and early 2000's, together with some of their then-rivals like Dr. Web. There's a reason they were
97.
by orbital-decay 2mo ago
Wikipedia's source policy makes it nearly impossible to refer to anything that is not in the media, and any sensitive article has to use weasel words like this. Are you just noting the issue with the article, or actually doubting that
98.
by orbital-decay 2mo ago
Bulk bitwise operations in a memory array driven by an out-of-spec controller is a thing, but I'm not sure why they're using NAND for this and not DRAM like everyone else, and why it's so small (just 80M weights). And the pag
99.
by orbital-decay 2mo ago
This looks like a cursed mix of different Apple eras. And has a paid Linux distro inside, amazingly powered by Homebrew of all package managers.
100.
by orbital-decay 2mo ago
I did not claim it (another user did), but I think your reaction captures the overall sentiment ITT, which is taking perceived status quo as some fundamental thing and defending the worse solution instead of changing the status quo. If all
101.
by orbital-decay 2mo ago
Huh? None of these force single-instance by default
102.
by orbital-decay 2mo ago
If that's the only justification, a good window manager might turn out to be a lot more convenient. Terminals are great for CLI. CLI integrated into GUI/TUI is absolutely amazing , but 99% TUI apps do not work like that, they&#x2
103.
by orbital-decay 2mo ago
Philosophical determinism is pretty similar in spirit to physics/CS (roughly same cause = same effect, no free will/side effects), and what parent calls determinism seems to be neither. It's rather about natural language prom
104.
by orbital-decay 2mo ago
Suddenly determinism has other meanings than "same input leads to the same output". I'm still confused by this and not sure how it happened so easily, but it seems to be accepted by everyone now.
105.
by orbital-decay 2mo ago
There are separate session persistence daemons you can use instead of a full-fledged multiplexer - dtach, abduco, shpool, zmx, etc.
106.
by orbital-decay 2mo ago
I don't know, I have too many tools that would really benefit from using GUI (e.g. coding agents, process monitors) but are using TUI and are written in TS instead. Everything about TUI screams jank to me, even the old kind. Both inter
107.
by orbital-decay 2mo ago
TUI got back into fashion and it seems like everyone is trying hard to reinvent GUI in a really roundabout way, like hacking Braille to display graphs. I think the circle will be complete with the support for proportional fonts for readabil
108.
by orbital-decay 2mo ago
It might be sensitive to the prompt and tool definitions the harness provides, I can imagine it overusing the tool if being directly asked to resolve ambiguities. But in a straight Fable vs Sol comparison where I control the context and mak
109.
by orbital-decay 2mo ago
Yes. Perfectly unconventional, like megameters!
110.
by orbital-decay 2mo ago
>Similar to what others have noticed, and as I predicted 8 months ago, better models are requiring less ceremony to work effectively. It has nothing to do with model capabilities, it's a result of purposeful persistence training a
111.
by orbital-decay 2mo ago
Was it straight line or orthodromic distance?
112.
by orbital-decay 2mo ago
Anthropic changed their character training policy many times, and the name is probably separate from it anyway (besides the bits from the constitution etc). Name training usually comes last in all models, if at all, and it's pretty sha
113.
by orbital-decay 2mo ago
Entirely possible, it always leaked but it's particularly bad in most recent models. Actually almost all issues with creative writing in LLMs are artifacts of either instruction tuning, alignment training, or CoT RL and seeding (newer
114.
by orbital-decay 2mo ago
It's hard to optimize for both decision making and creative writing at the same time, no wonder one of them is degrading in newer models.
115.
by orbital-decay 2mo ago
It's... complicated. Yes, RL reward hacking makes it learn "bird language" and yes, reasoning traces can be misleading. However they also pretty clearly steer the final reply and not simply justify it, and can stay somewhat c
116.
by orbital-decay 2mo ago
If you read and generate enough vibecoded readmes and front pages you can spot one right away. Structuring, no overarching idea, sentence patterns, neo-...isms like "lives" and "lands", etc. It's pretty clear that i
117.
by orbital-decay 2mo ago
Current AI is at least a few orders of magnitude less efficient than it could be (as evident from biological spiking networks, e.g. human brain). At some point the labs put too much work and money into transformers and nearly abandoned fund
118.
by orbital-decay 2mo ago
The article is literally talking about swarm self-coordination, which is an emergent [1] property that preserves and compresses the state while running, otherwise it couldn't do what they claim it to do. [1] After a training-time nudge
119.
by orbital-decay 2mo ago
The ladder in humans is even longer and wider than that, it's roughly: evolutionary pretraining of a complex molecular robot -> generational knowledge transfer and compression by the "parallelized agentic swarm" aka societ
120.
by orbital-decay 2mo ago
You mean simultaneously, and of course they are separate in humans, just not temporally. The models are learning continuously, the problem is that this process is fragile and has to be carefully curated, that's why it's separated

More ›