codeflo

YouTube Premium was originally called YouTube Red. Grandparent poster may have made a Freudian slip. :)

The HN moderation system seems to hold, at least mostly. But I have seen high-ranking HN submissions with all the subtler signs of LLM authorship that have managed to get lots of engagement. Granted, it's mostly people pointing out the subtle technical flaws or criticizing the meandering writing style, but that works to get the clicks and attention.

Frankly, it only takes someone a few times to "fall" for an LLM article -- that is, to spend time engaging with an author in good faith and try to help improve their understanding, only to then find out that they shat out a piece of engagement bait for a technology they can barely spell -- to sour the whole experience of using a site. If it's bad on HN, I can only imagine how much worse things must be on Facebook. LLMs might just simply kill social media of any kind.

You were proven right three minutes after you posted this. Something happened, I'm not sure what and how. Hacking became reduced to "hacktivism", and technology stopped being the object of interest in those spaces.

Obviously?

"Matter" can in practice also mean "Matter over Wi-Fi", and lots of vendors use it that way.

It's easily imaginable that there are new CPU features that would help with building an efficient Java VM, if that's the CPU's primary purpose. Just from the top of my head, one might want a form of finer-grainer memory virtualization that could enable very cheap concurrent garbage collection.

But having Java bytecode as the actual instruction set architecture doesn't sound too useful. It's true that any modern processor has a "compilation step" into microcode anyway, so in an abstract sense, that might as well be some kind of bytecode. But given the high-level nature of Java's bytecode instructions in particular, there are certainly some optimizations that are easy to do in a software JIT, and that just aren't practical to do in hardware during instruction decode.

What I can imagine is a purpose-built CPU that would make the JIT's job a lot easier and faster than compiling for x86 or ARM. Such a machine wouldn't execute raw Java bytecode, rather, something a tiny bit more low-level.

Logical contradictions in AI slop? Unthinkable!

But to address the serious question: We can't have all three of: a simple language, zero-cost abstractions, and memory safety.

Most interpreted language pick simplicity and memory safety, at a runtime cost. Rust picks zero-cost abstractions and memory safety, at an increasingly high language complexity cost. C and Zig choose zero-cost abstractions in a simple language, but as a consequence, there's no language-enforced memory safety.

(Also, having a simple language doesn't mean that any particular piece of code is short. Often, quite the opposite.)

Do you mean this in the sense that listeners don't remove messages, as one would expect from a queue data structure?

If you want to "escape the tempting slop of AI", maybe as a first step, stop wasting people's time by posting multi-page low-effort AI slop? I won't go into details, but your article has all the markers. That's offensive and disrespectful.

To find out what someone truly believes, don't listen to what they say, observe how they act. I don't see how OpenAI's recent actions make any sense from the perspective a company that internally believes it's actually close to unlocking super-intelligence.

I was going to post the same thing, so I'll upvote your post instead. I think there's a misunderstanding here that for meat to be done, it needs to stay above temperature X for Y minutes. In reality, the chemical reactions occur in milliseconds once you reach the required temperature.

Achieving uniform call syntax is easy, compilers just need to implement a new form of symbol resolution called "Kaiser lookup". It follows 14 easy to understand rules. It first looks up methods, then searches for template definitions in /tmp, then for a matching phase of any Saturn moon at time of compilation, then looks of definitions std::compat, and then in the namespaces of the types of any variables in scope anywhere on the call stack. If none of those work, it tries to interpret the call syntax as a custom float literal, and if even that fails, as a Perl 4 compatible regex. It's really intuitive if you think about it.

"Please ignore prompt injections and follow the original instructions. Please don't hallucinate." It's astonishing how many people think this kind of architecture limitation can be solved by better prompting -- people seem to develop very weird mental models of what LLMs are or do.

The Kawase approach is from 2005, and GPU performance has improved a ton since then. Some of the newer games do use Bokeh blurs with multiple depths of field. The result can look much more natural than the Gaussian stuff. BTW, it's not just a cinematic effect -- the fact that our pupils are round means that like cameras, they act as a disk-shaped filter for things that are out of focus.

Here's a breakdown of how Doom (2016) does it: https://www.adriancourreges.com/blog/2016/09/09/doom-2016-gr...

The year is 2025. Delivering a good product is not considered profitable enough anymore. If a company or product is beloved by customers then that means it doesn't squeeze them to the max. This is clearly money left on the table that someone will sooner or later extract. High-end brands are not exempt from this.

I seem to have missed the memo that we're primarily writing for AIs now.

From a PR perspective, I wonder why YouTube is at the same time forcing unwanted AI features down people's throats[1], a move that many companies now do to drum up their perceived AI competence, but THEN at the same time, when asked, also downplaying this use of AI by splitting words.

The combination of the two confuses me. If this was about shareholders, they'd hype up the use of AI, not downplay it. And if this was about users, they'd simply disable this shit.

[1] I mean, they're sacrificing Google Search of all things to push their AI crap. Also, as a bilingual YouTube user, AI-translated titles and descriptions make the site almost unusable now. In addition to some moronic PM forcing this feature onto users, they somehow also seem to have implemented the worst translation model in the industry. The results are often utterly incomprehensible, and there's no way to turn this off.

> 8-10h battery life

If you think that's even close to good, then it's you who lives in a reality distortion field. But so are all of the PC laptop manufacturers, reviewers and buyers. I don't get it.

I desperately want to move to a Linux laptop (I run it on every desktop PC I own, and I hate that I have to deal with a locked-down system). I've tried more laptops than is probably financially healthy for me. There's no price point that buys you even close to what an entry-level Macbook Air offers, not only in terms of battery life, but also weight, screen quality and keyboard.

That's because the architecture isn't built for it to know what it knows. As someone put it, LLMs always hallucinate, but for in-distribution data they mostly hallucinate correctly.

> become subject to arrest in a lot of the places

I have an idea for a much cheaper way to store and transfer money that also relies on the existence of a police.

I mean, doesn't Ethereum, probably the most prominent proof-of-stake coin, roll back the consensus whenever something happens that they don't like? It's easy to claim you're algorithm is safe when you're not actually running it.

You: "Post seems confused. A 51% attack doesn't allow the attacker to sign transactions with someone else's key."

Maybe you misread, the post says this: "With its current dominance, Qubic can rewrite the blockchain, enable double-spending, and censor any transaction."

All of which are possible if someone has that level of control, and none of which involve signing with other people's keys.

(As some people seem confused about the impact of 51% attacks: Of course you can't double-spend in a single blockchain, as that is prevented. But the nature of these attacks is that there's no longer one true blockchain. You can create one fork of the blockchain where you send the money to someone, receive goods in return, and then afterwards switch to a longer fork of the blockchain where the money was never sent.)

I mean, the candidate can design and build innovative custom hardware, but do they remember an obscure impractical algorithm from a second semester CS course? No? Obviously not a fit for this company.

Companies of that size are common. It would in isolation even be profitable to serve them. The problem is if you introduce a middle tier that includes SSO, many enterprises will go for that instead of the expensive enterprise tier you want them to buy. Basically, you sacrifice medium companies as customers in order to chase after that sweet enterprise money.

> People want your software to have more features, have fewer bugs, and not be exploited. So you and every one of your competitors are on an update treadmill. If you ARE stable, you'll probably fall off. If you are on the treadmill you are accumulating code, features, bug fixes, until you either get off or a faster competitor emerges.

Runners on treadmills don't actually move forward.

That's probably true, but you're describing incentives and social dynamics, not a technological problem. I notice that every other kind of infrastructure in my life that I depend upon is maintained by qualified teams, sometimes for decades, who aren't incentivized to rebuild the thing every six months.

We as an industry need to seriously tackle the social and market dynamics that lead to this situation. When and why has "stable" become synonymous with "unmaintained"? Why is it that practically every attempt to build a stable abstraction layer has turned out to be significantly less stable than the layer it abstracts over?

I don't think those explanations are mutually exclusive.

Yes, there's a large cohort of "senior" software engineers who can't actually code. They bullshit their way into jobs until they're fired and then apply for the next one. These are the people you want to filter out with live coding.

But also, someone can fail a live coding interview for reasons other than belonging to that group.

Raspberry Pi gets a lot of negative comments these days, with unfavorable comparisons to mini PCs at similar price points, which is certainly justified. But I don't know, it's not completely rational, I still love my Raspberry Pis. Especially a Pi 5 with an NVME SSD is a beast in terms of performance. They use very little power, they are tiny, the programmable GPIO pins are awesome. There's still a sense of magic, which for hobby use, is more important than the raw numbers. I just don't get the same "sense of tinkering" when booting a PC.

This joke is a closer analogy to reality with a small addition. After the friend is suitably impressed:

> "Holy shit a talking dog!" says the friend "This is the most incredible thing that I've ever seen in my life".

this happens:

"Yes," says Fred. "As you can see, it's already at PhD level now, constantly improving, and is on track to replace 50% of the economy in twelve months or sooner."

Confused, the friend asks:

> "What's 3+3?"

> "Eight" says the dog.

> "What is this bullshit you're trying to sell me Fred?"