Hacker Newsnew | past | comments | ask | show | jobs | submit | re-thc's commentslogin

> they are setting up in Singapore as their base

Europe in general has been tightening up their rules / taxes / laws around startups / companies especially tech and remote.

It's been less friendly. these days.


Yann Le Cun litteraly said this morning on the radio in France that it is headquarted in Paris and will pay taxes in France. Go figure…

No he said something like “well yes, only for the parts of profits made in France”

Why would it be any other way?

French people have this pipe dream all others french people to pay 75% of what they produce worldwide to pay for their retreats, hospital, useless schools system and all theirs “comité Théodule”

For such companies, France also offers generous R&D tax credits (Crédit Impôt Recherche): companies can recover roughly 30% of eligible R&D expenses incurred in France as a tax credit, which can eventually be refunded (in cash) if the company has no taxable profit.

Is that alongside 100% of R&D expenses amortized in taxes when a company has taxable profit covering them?

Yes indeed, if the company is profitable.

Doesn’t he live in New York himself? Although not sure if that matters depending on his role

There will be no corporate taxes for a long time, so alls good.

This is a singaporean news article from a singporean company[0] (Had to look it up)

As such, They are more likely to talk about singapore news and exaggerate the claims.

Singapore isn't the Key location. From what I am seeing online, France is the major location.

Singapore is just one of the more satellite like offices. They have many offices around the world it seems.

[0]: https://www.sgpbusiness.com/company/Sph-Media-Limited


> Europe in general has been tightening up their rules / taxes / laws around startups / companies especially tech and remote.

Like? Care to provide any specific examples? "Europe" is a continent composed of various countries, most of which have been doing a lot to make it easier for startups and companies in general.


> but an 5x efficiency gain in a single generation feels like its too much, especially considering how newer process nodes have been yielding less and less improvements

The efficiency is in other areas too e.g. memory, network, etc. It's TOTAL.

> Here's a synthethic benchmark page listing every GPU in recent memory

We don't have the GPU gains not because of process nodes. Nvidia and later AMD stopped investing in that direction. They started optimizing for AI not graphics.


> That's a tautology. People think chinese models are 10x more efficient because they're 10x cheaper

They do have different infrastructure / electricity costs and they might not run on nvidia hardware.

It's not just the models.


Except there are providers that serve both chinese models AND opus as well. On the same hardware.

Namely, Amazon Bedrock and Google Vertex.

That means normalized infrastructure costs, normalized electricity costs, and normalized hardware performance. Normalized inference software stack, even (most likely). It's about a close of a 1 to 1 comparison as you can get.

Both Amazon and Google serve Opus at roughly ~1/2 the speed of the chinese models. Note that they are not incentivized to slow down the serving of Opus or the chinese models! So that tells you the ratio of active params for Opus and for the chinese models.


AWS and GCP both have their own custom inference chips, so a better example for hosting Opus on commodity hardware would be Digital Ocean.

And Microsoft's Azure. It's on all 3 major cloud providers. Which tells me, they can make profit from these cloud providers without having to pay for any hardware. They just take a small enough cut.

https://code.claude.com/docs/en/microsoft-foundry

https://www.anthropic.com/news/claude-in-microsoft-foundry


> Both Amazon and Google serve Opus at roughly ~1/2 the speed of the chinese models

We were responded about 10x not 0.5x.

x86 vs arm64 could have different performance. The Chinese models could be optimized for different hardware so it could show massive differences.


These providers do not run models on CPUs, x86 vs. Arm is irrelevant.

They run Nvidia and Huawei for example. And mine was just an example.

Deployments like bedrock have no where near SOTA operational efficiency, 1-2 OOM behind. The hardware is much closer, but pipeline, schedule, cache, recomposition, routing etc optimizations blow naive end to end architectures out of the water.

Do you have evidence for any of this, or are you repeating a bunch of buzzwords you’ve heard breathlessly repeated on Twitter?

Evidence?

I mean GN has covered the Nvidia black market in China enough that we pretty much know that they run on Nvidia hardware still.

How is this related to the inference, may I ask? Except for some very hardware-specific optimizations of model architecture, there's nothing to prevent one to host these models on your own infrastructure. And that's what actually many OpenRouter providers, at least some of which are based in US, are doing. Because most of Chinese models mentioned here are open-weight (except for Qwen who has one proprietary "Max" model), and literally anyone can host them, not just someone from China. So it just doesn't really matter.

I mean sure, but in terms of cost per dollar/per watt of inference Nvidia's GPUs are pretty up there - unless China is pumping out domestic chips cheaply enough.

Also with Nvidia you get the efficiency of everything (including inference) built on/for Cuda, even efforts to catch AMD up are still ongoing afaik.

I wouldn't be surprised if things like DS were trained and now hosted on Nvidia hardware.


> unless China is pumping out domestic chips cheaply enough

They are. Nvidia makes A LOT of profit. Hey, top stock for a reason.

> I wouldn't be surprised if things like DS were trained and now hosted on Nvidia hardware

DS is "old". I wouldn't study them. The new 1s have a mandate to at least run on local hardware. There are data center requirements.

I agree it could still be trained on Nvidia GPUs (black market etc), but not running.


> The new 1s have a mandate to at least run on local hardware.

They do? Source?

But if that's true, it would explain why Minimax, Z.ai and Moonshot are all organized as Singaporean holding companies, with claimed data center locations (according to OpenRouter) in the US or Singapore and only the devs in China. Can't be forced to use inferior local hardware if you're just a body shop for a "foreign" AI company. ;)


> with claimed data center locations (according to OpenRouter) in the US or Singapore and only the devs in China

They just have a China only endpoint and likely a company under a different name.

Nothing to do with AI. TikTok is similar (global vs China operations).


> What AI are eroding is copyright.

At the moment it's people that are eroding copyright. E.g. in this case someone did something.

"AI" didn't have a brain, woke up and suddenly decided to do it.

Realistically nothing to do with AI. Having a gun doesn't mean you randomly shoot.


> This feels sort of like saying "I just blindly threw paint at that canvas on the wall and

> He fed only the API and the test suite to Claude and asked it

Difference being Claude looked; so not blind. The equivalent is more like I blindly took a photo of it and then used that to...

Technically did look.


The article is poorly written. Blanchard was a chardet maintainer for years. Of course he had looked at it's code!

What he claimed, and what was interesting, was that Claude didn't look at the code, only the API and the test suite. The new implementation is all Claude. And the implementation is different enough to be considered original, completely different structure, design, and hey, a 48x improvement in performance! It's just API-compatible with the original. Which as per the Google Vs oracle 2021 decision is to be considered fair use.


did he claim that Claude wasn't trained on the original? Or just that he didn't personally provide Claude with a copy?

I recon the latter, how would he know what was in Claude's training data?

> What he claimed, and what was interesting, was that Claude didn't look at the code

Who opened the PR? Who co-authored the commits? It's clearly on Github.

> Blanchard was a chardet maintainer for years. Of course he had looked at its code!

So there you have it. If he looked, he co-authored then there's that.


If I put my signature on Picasso painting, it doesn't make me co-author of said painting.

Blanchard is very clear that he didn't write a single line of code. He isn't an author, he isn't a co-author.

Signing GitHub commit doesn't change that.


> Blanchard is very clear that he didn't write a single line of code

He used Claude to write it. Difference? The fact that I write on the notepad vs printed it out = I didn't do it?

> Signing GitHub commit doesn't change that.

That's the equivalent of me saying I didn't kill anyone. The fingerprints on the knife doesn't change that.


I'll take a commit authored by someone else and then git amend the author to myself, did I write that commit then? By your logic I did apparently.

> I'll take a commit authored by someone else and then git amend the author to myself, did I write that commit then

I did say co-author didn't I? Even if you added 0.000000001% to something you did so technically, yes.

> By your logic I did apparently

If you take someone's email and forward it did you write that email? Instead of debating that imagine you took a trojan email and forwarded it to someone and they opened it - do you think you'd be held up in any way?


> because a great deal of writers have described the process of writing in detail

And that's often inaccurate - just as much as asking startup founders how they came to be.

Part of it is forgot, part of it is don't know how to describe it and part of it is don't want to tell you so.


> but we created MCP. Now we're wrapping MCP into CLIs...

Next we'll wrap the CLIs into MCPs.


> One CTO said their board explicitly asked "why do we need humans writing code when agents ship faster with zero complaints"

Because all the humans were fired?

Zero complaints where? News articles have been showing plenty of AI-generated issues. We've got more downtime than ever. There's been security breaches and all sorts of things.

Yeah why...


5.2 and 5.2 Codex is arguably the same gen.

Sure, but one is fine-tuned for what they are testing and one is not.

> but as the cost of fixing them trends towards $0

It’s not. In a proper org the cost is the testing, the release process, the coordination, the planning, etc.

Any scope creep even if it fixes something often gets shouted at.


AI can take over testing and release planning / coordination. This is the allure of AI. Being able to fully close the loop of releasing software without needing a human.

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: