Hacker News Reader

fspeechon Apple Used China to Make a Profit. What China Got in Return Is Scarier

If you are referring to Apple's sending engineers to work out problems and enforce standards you should know that this is not unique to Apple. Huawei does the same with its own suppliers. In my opinion it has much more to do with the fact that Apple and Huawei can afford to pay their engineers much better and hire people with good problem solving skills. It is not that they have some secret sauce and have to teach the factories how to make things. That would be a strange interpretation.

fspeechon Apple Used China to Make a Profit. What China Got in Return Is Scarier

Except it doesn't make any sense. What did Apple know about making phones? Back then the top dog was Nokia. It was FoxConn helping Apple figure out how to make a phone.

fspeechon Chinese e-commerce sites offer discounts of up to $351 on Apple's latest iPhones

Well the parts of the world that don't welcome Chinese competition deservedly pay a higher price for their phones. That is just the market at work. Apple isn't subsidizing their Chinese sales any more than Toyota is subsidizing theirs.

fspeechon U.S. and China agree to slash tariffs for 90 days in major trade breakthrough

But Trump already did a lot for them. What China fears the most is isolation, however unlikely that is.

fspeechon Sen. Cotton's bill would require AI chips to track location, curbing Chinese use

One version I read is based on ping time.

fspeechon Trade War Exposes China's Dependence on U.S. for Auto Chips

They exempted pretty much everything except for memories. Once designed in semiconductors are not easy to replace, but new designs will likely design these out as much as possible.

fspeechon Zhaoxin's KX-7000

They work at companies or universities, and there is a "market" to look to for the pay scale.

fspeechon Zhaoxin's KX-7000

Do governments allow some of their employees to be highly compensated relative to others? Would someone with real expertise in chip development work for the government at what the government is willing to pay? I think the answer is no.

fspeechon When Tech Bans Stimulate More Than Stymie China

"as opposed to"?

fspeechon China Halts Critical Exports as Trade War Intensifies

Your conclusion isn't logical. US per capita income is far higher meaning its economy is far more leveraged. To put it simply Costco employees need to push through way more goods per person, Google employees way more ads per person to sustain their much higher income. On the other hand, Chinese unskilled migrant export factory workers, while low wage, have a rural home to go back to and their standard of living won't change that much, while their skilled technicians can not be replaced en mass anywhere in the world. There's no way you can win a contest of pain against China. Think of the comparison as throwing sand into gears that are running at different speeds. You do far more damage to the one running at high speed.

The reason I brought up the NYT article is to say that even if you think change is important doing it rapidly is going to put the people that the policy is supposed to help through the wringer.

fspeechon China Halts Critical Exports as Trade War Intensifies

Isn't creating demands the US strength over China? Now you are saying it is not important for the US but important for China? Either way NYT recently did a good piece on the "China Shock" and the main takeaway is that people couldn't react to changes that happened too fast but they finally did. Now suddenly they are pushed to go the other way overnight.

fspeechon China Halts Critical Exports as Trade War Intensifies

You underestimate the importance of imports as a lubricant to the service economy. Sure you don't need the next new gadget, but many service jobs depend on your wanting it.

fspeechon Trump's Tariffs Wipe Out over $6T on Wall Street in Epic Two-Day Rout

Housing and healthcare are the two things that have little to do with trade. Housing affordability is almost entirely policy driven. In a "free" market (quoted because the definition typically doesn't take into account the necessity of a functional society to maintain property rights and values, just ask Detroit), housing affordability is entirely dependent on one's relative income and propensity to spend, i.e. you can have high absolute income (say in NYC) and still find you desired housing unaffordable. I think people intuitively understand this. Healthcare costs are more complicated but is also almost entirely a domestic issue.

fspeechon US Administration announces 34% tariffs on China, 20% on EU

That's supposedly done with the agreement of the creditors. There's never paying down of national debt over an extended period of time. That's not how modern finances work. Why they feel the need of cramming down friendly creditors is beyond me.

fspeechon US Administration announces 34% tariffs on China, 20% on EU

Exempting chips alone doesn't make sense. Does Intel have to pay 20% on ASML lithography machines or more on Japanese ones? Meanwhile TSMC can fab chips in Taiwan and export to the US without tariffs.

fspeechon US Administration announces 34% tariffs on China, 20% on EU

That runs against basic financial reality. The treasuries are the basis of all U.S. liquidity. Owning commercial papers or stocks doesn't help as companies own treasuries. Putting deposits in banks won't help either because banks hold treasuries.

fspeechon China, Japan, South Korea will jointly respond to US tariffs

Why is it a bad thing if they now like each other better (not sure if there is real evidence for it but assume it is true for argument's sake)? If American policy is what keeps them apart, which I strongly doubt, then it is a good thing that we butt out and let them integrate as much as they want.

fspeechon China Building Solar Station in Space Could Generate Practically Endless Power

How do you keep it stationary? How do you keep it out of the Earth's shadow a.k.a. night? How do you keep the transmission loss low and receiving station footprint small? How do you avoid harming things that could come into your beams? How do you achieve all the above simultaneously?

fspeechon China is developing some startling new kit in its quest to invade Taiwan

It shows up so naturally that one may think it is not whataboutism at work, even though the reasoning is often not well argued. We could choose to see it not as a moral judgement, but as questioning the justification of efforts pushing for conflicts with China.

fspeechon China is developing some startling new kit in its quest to invade Taiwan

They already can freely travel to and work on the mainland.

fspeechon China announces retaliatory tariffs on some Canada farm, food products

Unlikely as Canada ultimately depends on the US as a security provider, but yes, like Vietnam, it makes sense to trade with both sides. It's also why the US doesn't like it.

China's tax on rapeseed oil is like Canadian tax on EV, as China imports very little from from Canada due to prior trade frictions. But its rapeseed imports almost exclusively depend on Canada.

fspeechon China says it is ready for 'any type of war' with US

It may start sooner than you think. Now we will see how Missouri attempts to collect their $24 billon default judgement against China from a federal court for treble damage of "$8.04 billion in direct lost tax revenue" due to China's hoarding of PPEs. https://ago.mo.gov/attorney-general-andrew-bailey-secures-hi...

fspeechon China announces retaliatory tariffs on some Canada farm, food products

100% tax on EV imports and 25% metal tax imposed on China by Canada about half an year ago. Canada is in a tough spot. It needs to sync up its tade policy with the US in order to maintain its free trade with the US.

fspeechon DeepSeek's Multi-Head Latent Attention

Matrix absorption is unnecessary. What is needed is the order of multiplication associates towards the direction of the absorption. This and the modified Rope are needed to make the caching work.

fspeechon AI firms follow DeepSeek's lead, create cheaper models with distillation

The drive so far has been to scale up to push performance, not to reduce cost. Although everyone offers cheaper versions of models they don't base their marketing on these. Now that the momentum of scaling up on training is slowing down they are pushing inference time computes and efficiency comes to the forefront.

fspeechon AI, China and Copium

https://github.com/deepseek-ai/open-infra-index/blob/main/20...

Statistics of DeepSeek's Online Service All DeepSeek-V3/R1 inference services are served on H800 GPUs with precision consistent with training. Specifically, matrix multiplications and dispatch transmissions adopt the FP8 format aligned with training, while core MLA computations and combine transmissions use the BF16 format, ensuring optimal service performance.

Additionally, due to high service load during the day and low load at night, we implemented a mechanism to deploy inference services across all nodes during peak daytime hours. During low-load nighttime periods, we reduce inference nodes and allocate resources to research and training. Over the past 24 hours (UTC+8 02/27/2025 12:00 PM to 02/28/2025 12:00 PM), the combined peak node occupancy for V3 and R1 inference services reached 278, with an average occupancy of 226.75 nodes (each node contains 8 H800 GPUs). Assuming the leasing cost of one H800 GPU is $2 per hour, the total daily cost amounts to $87,072.

fspeechon AI, China and Copium

In addition Deepseek just showed that they are serving all that demand (admittedly inadequately) from a cluster of only 2000 H800 GPUs.

fspeechon DeepSeek-V3/R1 Inference System Overview

Gross margin in theory. They are using API pricing to project revenue but not all traffic is API. They don't charge for chat traffic so this is theoretical. It is in response to a raging debate on whether their gross margin is negative with their API pricing. Apparently not if one can saturate the traffic.

fspeechon DeepSeek-V3/R1 Inference System Overview

There are a lot of shared prefixes from user prompts. You can save by first looking into the cache to find the longest prefix for a prompt. Their MLA makes the KV cache particularly efficient.

fspeechon DeepSeek Open Source Optimized Parallelism Strategies, 3 repos

It's probably more about the size of the firm. Once a company is large enough with middle layers and silos it seems some form of KPI is inevitable. This could also explain why Deepseek doesn't mind open sourcing as they have not raised outside capital for expansion so they can't really scale rapidly on their own.