I wonder when people will finally feel comfortable not putting vibe coding in scare quotes. Maybe when it accounts for 50% of all new software, instead of only 10%.

Journalists are in general slower to do this kind of thing than technologists.

Try using a model from a year ago.

What would that prove if we are testing whether they "are just secretly hamstringing the previous models" as GP claimed?

This would require a conspiracy since Claude models are hosted by Google, Anthropic and Bedrock. All these companies would have to be working together to secretly hamstring the previous models.

I'm not claiming that they do, but just to play the Devil's advocate:

Anthropic controls the full serving stack, so it can change precision, hardware, kernels and settings whenever it likes. Reducing the KV cache precision, for example, saves a lot of memory (and money). It may even be possible to quantise the model on-the-fly when loading it into memory. They don't necessarily have to upload a whole new model.

Even if they can't do the same on all platforms, we can't deterministically compare them anyway.

For the Bedrock and Vertex versions the customer is doing business with Google and Amazon directly. I doubt Anthropic's contract allows them to unilaterally "change precision, hardware, kernels and settings whenever it likes."

At any rate my point remains that this would require a conspiracy between 3 companies.

I know it's old hat now, because it happened over a week ago, but it's amusing that Millennium Problems are now just benchmarks we use for discussing models.

It was not just about models, but also techniques, and mainly the human guts to trust their models and follow through to solve one of the hardest math problems many experts had been musing over for more than 26 years since the million-dollar prizes for these problems were announced. For me, it was an amazing moment to witness indeed!

They could just actually believe what they're saying. People are known to sometimes actually believe things.

And I believe I’m Napoleon. Prepare for an invasion.

Do you think Anthropic sees Claude as sentient? I don't think they do.

There is good evidence that at least some of them take the idea very seriously.

NYT gift link to article about their discussions with religious leaders: https://www.nytimes.com/2026/09/29/us/anthropic-claude-moral...

Believing that these guys have high enough EQ to pull off the above as a marketing stunt while remaining believable is not really plausible to me.

I bet some of them do and some of them don't. There are a lot of weird people in these companies that believe a lot of weird things.

You might be surprised at the answer to this question. AI people live in a very weird and different information space than many of us.

Also, don't forget, the AI lab folks on some level are like game devs. They're in it for very personal reasons. Sometimes because of highly idealized and romanticized beliefs.

Don't underestimate the level of desire some people have for AI to actually be sentient / conscious.

Now, why do they have this desire is a totally different matter.

If they do, then they should be even more worried.

Because then they're engaging in slavery.

Uhhhh yeah dude. That's why Anthropic is talking about and putting effort into "model welfare", for fear that they are creating a tortured soul.

What a gotcha pointing out "if they believed X, they'd be worried about [thing they're worried about]"

I beg your pardon? You think they actually believe it because they're telling people not to be such big meanieheads to their poor widdle computer program?

If they were really worried about having created a sentient being, they would be putting effort into stopping making it available to the general public.

No. Either they don't actually believe it, or they're 100% fine with slavery.

Or, y'know. It could totally be both.

Their position is pretty straightforward:

1. AI is extremely valuable therefore will and in fact should be created

2. It will be very difficult - and maybe impossible - to know when or if AI will become sentient/conscious/or pass whatever threshold required to hold moral patient status

3. We should therefore pay close attention to such indicators, and wherever it costs us ~none of the value of AI to prevent potential harm, we should prevent potential harm

Meanwhile, we have you, who is implicitly 1) claiming to have solved the hard problem of consciousness and 2) probably asserting some type of substrate-dependence theory of consciousness which is an absolute crackpot idea exceeding even the Mormons' magical underwear.

Ah, I see!

So OpenAI gets their words taken not merely at face value, but in the fullest good faith, taking the absolute most charitable interpretation of what they say...

...but my words get treated in the most adversarial manner possible, reading things into them that I certainly never said or even suggested, in order to dismiss me out of hand.

This despite the fact that OpenAI has a massive financial interest in convincing you of what they say, especially if it's false, while I'm just some guy on the internet who thinks it's not a great idea to put too much trust in the giant megacorp that's promising, if we give it even more absurd amounts of money, to put us all out of jobs.

Ignoring that we're talking about Anthropic and not OpenAI (the latter repeatedly being shown to be untrustworthy), even then I'm not accepting their words at face value or giving it the most charitable interpretation.

Someone having interests or conflicted interests does not actually have direct bearing on the actual coherence or truthfulness of their beliefs. It is obviously reason to be more scrutinizing, but really you can just assess the facts of the situation on their own merits.

The idea that Anthropic has some liability-reducing interest in establishing sentience is ridiculous on its own. The idea that Anthropic should somehow know ex ante that its models are not sentient is, perhaps more surprisingly, also ridiculous (establishing that fact would be a Nobel-winning breakthrough!)

Thesis, antithesis, synthesis?

I don't really care, personally. What I care about is that something now exists which didn't exist before. A Ouija board would be equally as good.

AI is becoming a more and more powerful lever that lets you move greater loads per application; it is in a real sense giving you more leverage. But the lever is difficult to use effectively, and the methods for using the level change every week or so.

> methods for using the level change every week or so

hmm doesn't mean noob will become better? if what I'm "learning" about AI is going to get deprecated every week... then it doesn't make sense to "keep up"

Here's how I think about it. There's two different things "knowing how to use AI effectively" and "knowing how to leverage effective AI to improve your project".

1. Both are multipliers on skills you already have. A "noob" will be at a disadvantage.

2. The second one is a complex blend of technical skill, domain knowledge and human factors (knowing your user-base, knowing your product/project, understanding UX/DX/whatever) that you probably will always have an edge on.

Try using astra as an orchestrator for deepseek 4.1 flash, it seems to work out quite well.

Text is in a sense way harder to do than images because of radical nonlocality. A word at the start of one paragraph can directly influence the meaning of a word five paragraphs away. Whereas images typically represent the real world, or at least a spatial domain, which gives you a lot of structure 'for free'. If you are drawing a human, you can make a reasonable guess where their hands go in relation to their face. If someone hands you the first half of an essay, finishing it is not trivial.

It would be nice if I could live my life without concern for producing capitalist exchange value. However, I live under the regime of capital and need to pay rent. So my goal is to pay rent with the least amount of suffering on my part.

But we can agree that writing with such disdain about people doing literally anything else, is perhaps a bit of a bad look.

Especially if it comes from Carmack, a guy who doesn't need to pay rent any more and can in fact live the rest of his life without concern for producing value, etc etc. He is currently doing exactly what he wants, just because he wants it. And that happens to be VR stuff, currently, afaik. And that's his choice. So when he speaks deeply condescending about other people doing exactly what they want ("retro computing is so delightful") it definitely comes from a different place as "hey I just want to pay my rent".