selcuka

> Over the coming weeks, we will add an invisible watermark to eligible ChatGPT and Codex text output in the European Union.

https://openai.com/index/eu-text-provenance/

It depends on how you use your subscription. If you are "vibe coding" and spinning up a lot of subagents in parallel, a $10-20 subscription won't be enough. If you usually open one or two terminals at a time, and use it like you are peer programming with it, then Anthropic (Opus 5.5) or OpenAI (Sol 6.1) are the best $20 subscriptions.

ChatGPT output is watermarked, but I guess the "invisible unicode" trick is easy to circumvent, and TextGrain is only activated for EU users.

People executed in the first half of 2025 spent an average of 24 years on death row. In practice, it's a life sentence plus execution. I believe the $1-3M figure is on top of this.

But they were already paying for the crimes of an individual, in the form of time wasted, and parking fees. Their suggestion was about minimising it.

Can't you install your own CA to Android phones?

> We would usually wash them anyway when we arrived because I’m crazy.

We wash the dishes (and sheets, if there is a dryer) when we arrive, and descale the electric kettle and the coffee machine. I guess I'm crazy too.

True. I'm not aware of a $500K+ bonus system for "engineers, support staff, lawyers, and some supporting roles like finance and security/compliance" in America.

Sure, but remember that Verisign manages a total of 179 million domains. They also have less than 1,000 employees. Can they really have a salary cost of >$530K per employee? It's obviously the executive team that pockets most of that cost.

> Is there a directory of private equity acquisitions?

CrunchBase is usually accurate:

https://www.crunchbase.com/organization/namecheap#financials

> But we are talking about saving 10$/year? That's less than 1$/mo.

I agree in principle, but businesses commonly own more than one domain name, so the costs add up.

Peasants don't pay for the subscription. The employer does.

Ollee watch [1] is compatible with F-105W:

> Tested: All Casio A168 and F-105W models that use module 3298 (manufactured since ~2012).

[1] https://www.olleewatch.com/shop/p/ollee-watch-one-kit-syp37

Would you call Qwen 3.6 35B "similar size" with ~50% more parameters (both active and total) though? I believe what GP meant that if there was a Qwen 3.8 24B-A2B, it would roughly be as intelligent as this one.

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.

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

Apparently Samsung Galaxy S26 Ultra has a 200MP camera with a 16,320 × 12,240 pixel resolution. So in a few years phones might be able to take pictures which can't be converted to WebP.

The fact that there is a size limit (16,383 x 16,383 pixels) is still annoying for some (rare) use cases.

Google bots definitely come disguised, to check whether the web site serves different content to non-Google user agents.

It's even in the Urban Dictionary [1]:

> Linux ISO: A codeword for copyrighted material shared without permission, usually over P2P networks.

[1] http://linux-iso.urbanup.com/2782626

> why selling 1 high-efficiency AI chip when you can sell 10 low-efficiency GPU ?

Because you can't use the former for anything else? You will have to buy a new chip every time a new model is released.

Not to mention that it will be much more expensive.

There's an xkcd (1172) for that.

> AMD announced their acquisition of Taalas. That doesn't exactly inspire confidence that their frontier LLM will arrive as promised.

Why not? AMDs chip design experience and production capacity are magnitudes larger than a small startup. Assuming that they acquired Taalas for their technology, I don't see a reason why they couldn't.

Also "faster" indirectly means "more intelligent" with reasoning models.

For example, Opus 4.6 Max was somewhere between Opus 4.7 Medium and High in some benchmarks, but it was slower. If there was a way to run it 10x faster, the economics would be different.

I'm wondering how many automated tests this change broke.

Yes, I know it's not the example.com's fault, and it's a side effect of Hyrum's Law:

> This is not a service, avoid relying on it for testing and monitoring purposes.

Don't we all have a few fragile tests?

They can actually transplant a single liver to two people by splitting it [1].

> Approximately 1 in 10 donated livers can be divided into two segments or grafts (this is known as split liver). [...] By splitting the liver into two segments, the donated liver can be used to help two people waiting for a liver transplant.

[1] https://www.nhsbt.nhs.uk/organ-transplantation/liver/is-a-li...

> he needs to also donate it to a third person

It's more difficult to do it while you are alive when you have a single kidney.

It's not complete nonsense. The rejuvenating effects of young blood transfusions have been known for some time.

https://pmc.ncbi.nlm.nih.gov/articles/PMC5836258/