coldtea

4 ms·

1.
by coldtea 5d ago
The article claims a "nightmare" but offers little to support it. "Oh, I have to make some choices" like Codex or ChatGPT, or "Chat vs Work" is not a nightmare. Also, clearly AI written itself.
2.
by coldtea 5d ago
It very well may. Quite a few among those that have been fired because of AI (or at least, with the companies claiming AI as the cause).
3.
by coldtea 5d ago
> But what is that cobbler supposed to do during that time? Optimally? Move to the best available trade that appears not to be deprecated for a good 50 years in the future, and in which they'd hire someone of his background.
4.
by coldtea 5d ago
As an AI user you're not that different from the punch cards programmer. You're more like a cobbler who adopted some new electrical tools for their craft, and feel smug against backwards cobblers using manual tools, while in the m
5.
by coldtea 5d ago
Depends on the kind. 400 locs JS or Java or something is barely worth mentioning. Can do more, a lot of it is boilerplate bs. 400 locs of critical C code, sure.
6.
by coldtea 6d ago
The phrasing still makes it sound like it's due to the LLM. It's not. It's do to compiler and CPU, and OS optimizations, and a non-LLM program could suffer the same just as well.
7.
by coldtea 7d ago
For allignment maybe. For the core functionality and the optimizations we don't really need to know how the emergent capabilities decide on particular answers. Which is why we could build LLMs before those features ...emerged for us to
8.
by coldtea 7d ago
BS. Run them sequentially on a single core, and without fancy speedups enabled, and they do. The algorithm is determinstic. Any non-determinism present with 0 temperature it's not some mysterious LLM-inherent property, but something th
9.
by coldtea 7d ago
> But to be clear, nobody understands why the evolved weights of a NN make the decisions that they do. We might not understand particular "emergent" capabilities, but the low level mechanism is not just understood, but a deter
10.
by coldtea 10d ago
It's 2026, with cleaner high ISOs even in phone sized sensors giving the ability to raise the shutter speed as needed, we hadn't had much of an issue with blurry photos for a decade now, with or without OIS. There have been severa
11.
by coldtea 10d ago
> For other tinkerers, the tinkering is accomplishing an objective or building a thing Those are not tinkerers then, as there's no tinkering involved. Or there might be in other parts (e.g. with electronics to hook up and house the
12.
by coldtea 10d ago
"I asked somebody else to do this hobby project for me" is not the same as "I tinkered" whether that somebody else is a person or AI. Same how you're not a musician just because you had Suno create something. So, no
13.
by coldtea 10d ago
> But the assumption that smartphone cameras, including those used in 99USD smartphones, will become 100% cryptographic cameras in a few years is highly unlikely, considering that those cameras didn't even gain OIS in the last 13 ye
14.
by coldtea 10d ago
> There is universal value in OIS, as everyone takes pictures while holding the device in his hands, yet OIS is not applied universally in all price-tiers of devices. The reason is COST. The reason is digital stabilization is a good eno
15.
by coldtea 10d ago
> If COST is not a factor, why does the Galaxy A16 still have no OIS (Optical Image Stabilization)? Because it's just not that important, phones and cameras have also used digital stabilization via cropping since forever.
16.
by coldtea 10d ago
> There's no way to link a reference image to a person; it's also not possible to determine if a pair of images came from the same device. Apple knows the iphone the reference image was uploaded from, so, yes, there is.
17.
by coldtea 10d ago
People will go "no, I nerd out and tinker more than ever" where by "nerd out and tinker", and what they mean is "I prompt the AI, and it does the tinkering for me". I say the question isn't if you use AI t
18.
by coldtea 10d ago
It is the YAML spec's fault. The implementations get away with it because the spec is not strict enough, plus all the magic it uses it bad (like the types being guessed implicitly in some case, or NO treated as boolean).
19.
by coldtea 10d ago
Enough? They're just getting started...
20.
by coldtea 14d ago
They sell to dozens of major companies, plus sell their own retail. An added 0.4% revenue for selling the most expensive part (the display) of the new flagshit product, for the most popular high end phone line in the world, sounds about rig
21.
by coldtea 14d ago
Apple still has their 30% margins on top of whatever they give, like always. And has used Samsung and other parts since always. They, on the other hand, have razor thin, mass retail, margins, for what they sell to Apple.
22.
by coldtea 14d ago
So? It's an insignificant amount compared to Samsung/Apple scale revenues. Samsung's revenue for 2025 was 233 billion. They get 0.4% more. So?
23.
by coldtea 15d ago
Where exactly do I say "Apple _shouldn't_ add it to their earbuds"? I'm saying as an "incentive" to get to the new version of airpods 5, it's a lame one, since the feature existed in all kinds of cheaper e
24.
by coldtea 15d ago
> I’m more and more convinced that all of AI engineering is Neijuan (内卷, meaning curl inwards). In China it describes a system that demands ever more effort and competition without improving output. The way in which it sometimes shows up
25.
by coldtea 15d ago
Wild that you think he is not. He is clearly a product of a mindset and a large culture share, with a big cult following (way before he was a president) and he was voted into the highest office twice. > Has HN really fallen this low? W
26.
by coldtea 15d ago
> This has only ever been true in a relative sense. Until recently, one was expected to understand the software-level code they wrote. But not the machine code, or the computer internals at a deep level. Just because something wasn&#x27
27.
by coldtea 16d ago
> Up until now, LLMs have required outside input and interpretation to avoid collapsing under their own weight. This has been known for decades. What does "for decades" mean? That in 2000 we knew "LLM require outside inpu
28.
by coldtea 16d ago
> Did you guys ever manage to create a perpetuum mobile? Every time it is mentioned somewhere, it is fraud. An LLM should comprehend that it needs (trained) humans to exist, to evolve, and to be relevant in any metaphysical aspect of its
29.
by coldtea 16d ago
They're the same in this regard, and it’s insane that you think otherwise. Asians might get the fame today with movies like "Crazy Rich Asians" but Americans have long been seen as the typical "flaunting wealth" con
30.
by coldtea 16d ago
The iPhone 18 Pro and the Airpods are equally dissapointing. Even the folding iPhone is nothing special.

More ›