nsoonhui

I wonder why this is not guardrailed by Opus?

I fed a few pieces of my (anonymous ) writings to ChatGPT and asked it to guess whether it's me. ChatGPT refused, "due to policy to not doxx people".

I don't think this has much to do with export control-- note that Manus, as impressive as it is, is still a wrapper around fundamental western models--, rather it has more to do with capital controls.

China has been trying to stop large scale outflow of businesses and individuals for quite some time, due to local politics concern. What Manus was doing, achieving successes first in China then setup a nominal shell company in Singapore, seems like a textbook case of flight (润）, which China is trying to prevent.

I said the same thing as you and I got summarily downvoted (https://news.ycombinator.com/item?id=47888227).

That HN is quick to upvote an unsubstantiated comment ( the grandparent one, because it aligns with the anti US bias? ) and downvote fact finding one doesn't bode too well for the community as a whole. I have seen enough how polticial ideology colors everything in my home country( Malaysia), and the decline of the country is palpable, and I don't expect to find such a thing here. We are supposed to be impassioned and rational, right ?

Render to Jesus what's due to him, ditto for Caeser.

OK, strange that it doesn't appear on my version of the webpage

https://api-docs.deepseek.com/zh-cn/news/news260424#api-%E8%...

This is the first figure of the section that the above links point to (https://api-docs.deepseek.com/zh-cn/img/v4-spec.png).

And I can read Chinese.

Sorry, but exactly where did you get the idea that DS V4 runs entirely on Huawei?

I asked DS itself and it denied this. It says: 'Nvidia chips are absolutely used for DeepSeek V4. The reality is a pragmatic "both-and" strategy, not an "either-or."'

And based on the DS V4 technical report (https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main...), it is mentioned that:

  We validated the fine-grained EP scheme on both NVIDIA GPUs and HUAWEI Ascend NPUs platforms. Compared against strong non-fused baselines, it achieves 1.50 ~ 1.73× speedup for general inference workloads, and up to 1.96× for latency-sensitive scenarios such as RL rollouts and high-speed agent serving.

(In all honesty I relied on DS to give me the above, so I haven't vetted the information in full.)

It mentions that Nvidia is still used. It doesn't even mention that Huawei chips are used in production — only in testing and validation, yes.

Sorry, but exactly where in the article that you linked contains the mention of " Ascend 950"?

To me this is the perfect proof that

1) LLM is not AGI. Because surely if AGI it would imply that pro would do better than flash?

2) and because of the above, Pelican example is most likely already being benchmaxxed.

I think it's important to see that the other similar example, a dragon driving a car while eating hotdog, doesn't nearly render as well.

https://news.ycombinator.com/item?id=47865232

This is an astonishing claim and if true, will make AI a lot less useful in real life scenario.

In real life, take programming as an example, we want Claude to be strong in capability at first, but what is more important is for it to learn our code base, be proficient in it, as it gains experience around it. In other words, become a domain expert.

Because our code base is proprietary I don't expect ( not do I want) the AI to be familiar with it on the first day. So learning on the job is the only way to go.

Only in that way it will resemble a human programmer, and only then we can truly talk about replacing human programmer.

  it seem they are making good progress on their huawei ascend chips

This is interesting to me. I thought that the reason for deepseek delay was because of the insistence ( by the politicians) to use huawei chip[0]. But that was last year August.

Anything changes in between?

[0]: https://www.reuters.com/world/china/deepseeks-launch-new-ai-...

From my experience the biggest difference between AI and junior programmer is that, AI can churn out code very fast, but you need to do the testing and verify the fix. Junior, on the other hand, is very slow in writing code but can do the verification and testing on his own.

Usually the verification and testing is the most time consuming part.

I am working on graphical application like AutoCAD, for the context.

Because

1. Physical dongle tends to break, and when it does, they expect us to give them replacing parts

2. They do expect bug fixes-- especially calculation bug fixes-- as the bugs are discovered. It's hard to leave their production critical apps broken like that once you know that the bugs can cause monetary or even life loss.

In a perfect world, I would have agreed with you, even if it's diametrically opposite to my interest as a software developer cum business owner.

But in an imperfect world whereby our dependencies ( software components that we use) and platforms that we need to build/rely on ( like Civil 3D) do charge us on annual basis, and that some of users expect perpetual bug fixes from us, with or without a support contract of sorts, SaaS seems to only way to go for our sustainability.

I write civil engineering software [0] and am familiar with this kind of dongle. Yes, even today there are users who want this kind of dongle instead of, say, cloud-based validation. They feel secure only if they have something tangible in hand.

Since we sold (and still sell) perpetual licenses, it becomes a problem when a dongle breaks and replacement parts are no longer available. Not all users want to upgrade. Also, you may hate cloud licensing, but it is precisely cloud licensing that makes subscriptions possible and, therefore, recurring revenue—which, from a business point of view, is especially important in a field where regulations do not change very fast, because users have little incentive to upgrade.

Also, despite investing a lot of effort into programming the dongle, we can still usually find cracked versions floating online, even on legitimate platforms like Shopee or Lazada. You might think cracking dongles is fun and copy protection is evil, but without protection, our livelihood is affected. It’s not as if we have the legal resources to pursue pirates.

[0]: https://mes100.com

Last Thursday windows 11 forced this update on my Acer machine. It caused me BSOD: inaccessible boot device [0], so I had to reformat my machine to get Windows running again. You might think that in 2026 you shouldn't get BSOD, but here we are.

So I am now very wary of any Windows updates, including a Out of Band Update [1], which it is claimed that it resolves some issue. However, since it's never mention whether the Out of Band Update will solve mine, I'm very hesitant to update.

[0]: https://www.windowscentral.com/microsoft/windows-11/windows-...

[1]: https://news.ycombinator.com/item?id=46750358

Let's do a thought experiment and take it to the extreme: why not tax at the maximum?

We have already tried that in human history, it's called communism. No one is allowed to take private profit, everyone contributes to the best of their own ability, and everyone consumes according to their needs. It should be utopia because there is no wealth gap and wealth is maximally redistributed. Which is exactly what taxation is designed to do, only to the most extreme.

And I think everyone will agree with me that communism is a miserable failure. The rich may not leave physically but mentally they are checkout -- not willing to work as hard or take as much risk. So the answer is yes, if you tax them, most certainly they will leave physically for haven with lower tax, all things being equalled. Or leave mentally.

But not all things are equalled, so you can still tax them at a somewhat higher rate provided that you can provide other incentives. But still, too much tax will make it more likely for those who are able to to leave. This is almost an axiom.

Here's a similar discussion[0], and here's my experience[1]:

Last Thursday windows 11 forced this update on my Acer machine. It caused me BSOD: inaccessible boot device, so I had to reformat my machine to get Windows running again.

So I am now very wary of this Out of Band Update[2], especially when it's not mentioned whether the latest update solve my issue or not. I don't know the same problem is still there, or whether this update makes the problem any better or worse

[0]: https://news.ycombinator.com/item?id=46761061

[1]: https://news.ycombinator.com/item?id=46761870

[2]: https://news.ycombinator.com/item?id=46750358

>> "but why would you do that?"

My bread and butter is Windows WPF cum AutoCAD-like application. My users are all on Windows. So I have to develop on Windows.

Last Thursday windows 11 forced this update on my Acer machine. It caused me BSOD: inaccessible boot device, so I had to reformat my machine to get Windows running again.

So I am now very wary of this Out of Band Update[0], especially when it's not mentioned whether the latest update solve my issue or not. I don't know the same problem is still there, or whether this update makes the problem any better or worse.

0: https://news.ycombinator.com/item?id=46750358

Last Thursday windows 11 forced an update on my Acer machine.It caused me BSOD: inaccessible boot device. It took me a reformat to solve the issue.

I am now very wary of this Out of Band Update. I don't know the same problem is still there, or whether this update makes the problem any better or worse.

I turn on the `showdead` option, and at the `New Links` section I can see a lot more "Show HN" posts that are flagged dead.

I am basing on Terrence Tao comment here: https://news.ycombinator.com/item?id=46665168

It says that the OpenAI proof is a different one from the published one in the literature.

Whereas whether the Deepseek proof is the same as the published one, I dont know enough of the math to judge.

That was what I meant.

No it's not, you can refer to my link and subsequent discussion.

I am not familiar with the field, but any chance that the deepseek is just memorizing the existing solution? Or different.

https://news.ycombinator.com/item?id=46664976

It's not directly comparable. The first time writing the code is always the hardest because you might have to figure out the requirements along the way. When you have the initial system running for a while, doing a second one is easier because all the requirements kinks are figured out.

By the way, why does your co-founder have to do the rewrite at all?

This is interesting, how do you get it done? From what I know CAD tools generally don't support text file, only binary blob which is LLM unfriendly?

Do you consider adding support for AutoCAD or AutoCAD vertically integrated software like Civil 3D?

Sergey might have some positive influence on Gemini, but given that he isn't an AI scientist ( AKA no technical background), I really do wonder what sort of influence that (only) he could have had, beyond just bringing in key people.

When I started my company, we suddenly found that we were in a good small fortune, not enough to be millionaires or billionaires, but enough to get people to run the business semi automatically with very minimum input from the founders.

I took a semi retirement approach to the business, there really wasn't a lot of things to do, my role was sort of just "managing" programmers. I got so much free time that I could even start a second business on the side.

Despite my best ability to stretch my work, I couldn't even fill up half of my working hours. One would have thought that this is heaven. But the time I was most free was also the time I was most miserable. I wasn't happy, I was gaining weight, I was perennially asking myself why the business couldn't be bigger and I couldn't sell it, so that I can be real millionaires and billionaires with financial freedom!

Then fate intervened, the sudden fortune disappeared and I no longer had the luxury of just "managing people"; I have to do hands-on. And it was this activity, the feeling that I was contributing to something, that I was writing code again and actually building stuffs, that made me happy again.

Today we are bigger than what we once we were, but still, I am writing code and pretty much hands-on.I vow that I will never retire, even though if I could. Because it's the meaningful work that sustains life and provides happiness. Being able to work on it is a luxury that I will never want to give up, ever.

I assume that you are engaging in good faith, therefore I will respond.

On the meaning of Palestine will be free, don't westplain the Palestinians by reading your interpretation into their mind. Instead, listen to what they actually said.

https://youtu.be/w4iGFT9Yl9o?si=oWKWAUzlMSec4n67

A lot of misunderstanding about Israel stems from people not reading the situation as it is, ie: listening to what both sides actually say, instead, they are listening to their own projections of the Jews and the Palestinians.

And your take that on Oct 6 Israel held thousands of Gaza doesn't explain why Israel would unilaterally pull out from GAZA in 2005, which is just another way of saying that it's likely to be false.

Maybe Hi Tech weapon is impressive but that could lure us into false sense of security. Israel learnt the lesson the hard way in the October 7 attack.