unclebucknasty

Understandable. But it doesn't seem completely unreasonable in this case, given that AI labs have clearly exhausted much compute, the all-out effort to build data centers is underway, and memory / other resources have been strained for some time.

Are you deliberately performing "the thing" in your reply as irony, or have you found a PT Cruiser that doesn't hallucinate?

This has all started to feel very Emperor's New Clothes-ish. "Well, if it's not working for you, then it's your fault. You're just not using it properly. You're not worthy of its wonders".

Not to say there's no value to be had. However, it's also been oversold and its error-proneness is a real problem.

But, it seems there's a steady stream of commenters and creators who are quick to remind us that it's never a problem with the tech. It's always a user problem.

Meanwhile, the errors are renamed "hallucinations", but let's not focus on those. In fact, if you just look right over here at this .md file...

You previously asserted that devs repeatedly ignored specific code quality instructions, which is clearly a fireable offense. You're now generalizing out to a vague concept of "sloppy code", which is not even quantifiable.

In any case, the point is that low quality code shouldn't be promoted as the acceptable goal for AI.

>tbf many human devs I've worked with don't...

OTOH, if someone insists on writing bad code—even after being advised otherwise—then they should expect to be out of a job.

Doesn't seem like "terminated junior developer" is the level of quality we should be accepting or promoting for AI.

I still have to remind Ai not to use the n+1 pattern in queries, and fight it to stop using inconsistent inline styles vs classes, even within a single page.

How are people getting this "waste of time not to use AI" and "generates 99% of my code" level of code quality?

I know people's response will likely be something, something, "add instructions to memory to not use n+1", etc. But if the Ai wants to generate code this bad, then it represents an overall quality risk that would require an infinite memory file.

So, it's no more remarkable than what LLMs do every day, but more remarkable than the solve claimed in the article?

So, none of this is noteworthy, but we're...noting it?

Not quibbling with you, personally. Quibbling more with the state of the hype.

>Now it's weird if an agent isn't writing 99% of your code.

>But September 2028? It might just be weird to run companies without AI management.

Coding is pattern matching and still requires a human in the loop to manage.

But, any human in the loop managing the "Ai management" is the manager, by definition.

Only way true Ai management is viable is via non LLM Ai, so completely depends on advancements there.

>That's potentially more impressive...

Serious question: why? Is it not just pointing it at its massive training corpus for a list of unsolved problems, and possibly even by degree of perceived difficulty? I'm trying to understand why finding the problem isn't a simple "search engine" style challenge, at which LLMs excel?

That's exactly what it feels like and it makes perfect sense, given the known compute constraints.

What do you mean? As in, they are reallocating compute away from previous models?

If so, I would think that would result in worsened performance, not quality (unless you are also suggesting they may be quantizing).

You mean like quantized?

>I thought models are getting dumber, but benchmarks were convincing opposite

>Opus 4.8 and Opus 5 seems worse models than Opus 4.6

After all we've heard about benchmark cheating, I'm earnestly not sure which or whether benchmarks are reliable anymore. But, beyond the models, I wonder if changes to their harnesses and/or instructions dumb them down. I have noticed models change their behavior, even when using the same version/effort. Sometimes for better. Sometimes for worse.

And, I have noticed a model go from really good to struggling. On 4.8 things were going well for a good stretch, so I did not switch to 5 when it came out. Even after hearing complaints about 5, 4.8 was still going well. Then, suddenly over the last few days, 4.8 seems to have nosedived. It feels similar now to the complaints I hear about 5.

In my case, it suddenly started ignoring my design guide, and introducing new fonts etc. It would even use several different fonts and sizes, as well as different margins for similar elements within the same page. It abandoned classes and started inlining styles. It started feeling random and, even after it realized it needed to go back to the design guide, it just continued with more of the same.

There seems to be something that happens after new model releases in both quality and behavior of previous models. It may not be immediately, but eventually there is frequently some regression.

>It's an empty point and a boring one and leads nowhere.

But, is it true?

>It means nothing...shallow...

Ironically, the entirety of your comment just repeats that the GP comment means nothing. There is no further explanation or "depth".

Maybe true if the discussion was an explicit comparison between the U.S. and modern Europe.

As it is, no one was discussing Europe, let alone holding it up as a model society.

>This differentiates the US from who, exactly?

What-aboutism is not a very compelling argument.

I sometimes wonder what the tipping point is for people, or whether there actually is one for many.

It's as if they are looking for an exacting duplicate of previous regimes before they become concerned—the agency name must be "Stasi" and uniforms must be of the same color.

History doesn't repeat itself, but it often rhymes.

-Mark Twain

This is what I meant by "imposing specific choices, determined by a key". Maybe "impose" or "specific" were too strong in my attempt to simplify?

I attempted to clarify that the impositions themselves are not deterministic, by indicating that the entire process is still probabilistic.

Maybe Anthropic's explanation is simple enough [0]:

>When watermarking is used, choices are still made at random, but the source of the randomness is different. Instead of using an arbitrary random number generator to pick the next word, watermarking uses the key and a few words that come before to settle what word the model should pick. That is, the words that Claude picks are still random, but now, one can check the sequence of words and see if it’s consistent with the choices Claude would make if it was using the key. If it is, one can assign a probability that the text was generated by Claude.

[0] https://www.anthropic.com/news/claude-text-watermark

Simple version: In instances wherein the otherwise statistically chosen next word is a "toss-up", watermarking removes the randomness by imposing specific choices, determined by a key. This then becomes a detectable pattern when scanned with the key (stastically—detection itself is probabilistic).

>use it to store arbitrary information

No additional data is embedded. The range of available data is constrained by the text being generated (i.e. the sets of "next words" per text).

The style...

CC:

"The problem is that I overreached..."

[Wall of words here]

"Two things: window surface is limited. Extract template. Buffer result and add to surface. Then, follow-up with new model..."

Me:

What do you mean by "window surface" and what result are you referencing? Also, why do we need a new model?

CC:

"Ah, you're correct to point out that no new model is needed. The problem is elsewhere and once we address that, the existing model should work fine. Now, as to your question about..."

[Wall of words here]

I noticed a few releases ago a more marked shift to a kind of conversational shorthand that seems to be intensifying — using phrases instead of complete sentences, and its own style of jargon, wherein it introduces new terminology on the fly.

This is especially common when it is trying to explain an issue, what it's done or what it's proposing to do. I think the idea was for it to be more concise, but it's actually still verbose, only not written in complete sentences. So, it frequently reads as somewhat cryptic and requires rereading to parse.

So you have this wall of words, followed by an explanation that is harder to read and introduces new terms to reference something in that wall.

For example (not a copy / paste).

[Wall of words here]

"Two things: window surface is limited. Extract template. Buffer result and add to surface. Then, follow-up with new model..."

That's obvious nonsense, but it will be very much of that style. And some of the terms like "window surface" are a recharacterization of something in the wall of words that you must now tie together.

So, on first read it can have that same nonsense feel as above, and you have to really lock in to make sense of it.

You're running Cloudflare Pages on a home server?

What would really be crazy is if there were a platform on which people provided commentary about other people's commentary on other people's commentary.

To GP's point, any connected server can become a liability, even if it is intended to just host static HTML (or for some other purpose).

>Is it the AI or the number of cameras?

To be clear, the controversy didn't just arise with AI—it only intensified.

>we should pay a team to spend hundreds of man hours...

Some argue that none of it should be happening, so would say that it should be prohibitively expensive. But, again, the added "efficiency" makes the problem far worse. Inefficiency is the throughline of the Fourth Amendment.

You've made clear that you value efficiency. But, you've not addressed any of the actual arguments against it.

>There is zero privacy in public spaces

You've agreed that this fact does not preclude uncomfortable scenarios, even for yourself. So, the discussion is about where people fall in their comfort.

I would sum up my response to the rest of your comment by noting the difference between individual citizens recording as they please, and a state-operated, always-on surveillance network, with the capacity to surveil every citizen outside of their homes.

It's worth noting here that the government famously faces stricter constitutional limits than do citizens.

The inefficiency you decry is what others value, as it protects privacy and constrains state power. There's long been a trade-off between efficiency and rights, which much of our legal system seeks to balance—as enshrined in various amendments.

Abuse of the frankly violative sort outlined in this article is but one concern, and likely cannot be properly contained in any case—particularly absent the requirement for warrants. This is precisely the reason warrants exist. We don't merely trust law enforcement to abide by its own "rules".

So, the larger point is that the very existence of state-operated surveillance networks is considered an intrusive form of abuse by many. One that uses previously unimaginable technology to circumvent the spirit of the Fourth Amendment, which consistency tells us would have otherwise explicitly protected us from such abuse.

>maybe exposes hypocrisy

I don't know that hypocrisy is the word, but it does seem that you're going a long way to carve out an exception, where there is no real difference.

That is, it appears that you've simply decided one scenario is okay, even though that scenario itself begs the question. The article on which you're commenting demonstrates one reason that's so—the potential for abuse.

So, seems it all boils down to "out of sight, out of mind" in your framing. As long as you don't see the people invading your privacy, you're okay with believing that no one is. You don't mind that all of the tools of surveillance are in place, because you've defined a scenario wherein it's not possible that they're being used to actually surveil you.

In sum, your thought experiment is essentially, "would this be OK in a world that doesn't exist". But, you're answering "yes" in the world that does.

Maybe that's what you're sensing as "hypocrisy".