schneehertz
schneehertz on Hacker News
Japanese culture treats an apology as the end of an incident; afterward, they no longer feel guilty, nor do they continue to do anything about the mistake.
DeepSeek can accomplish over 90% of tasks of equivalent value at one-tenth of the fully loaded cost of O/A, and this capability can further spread through open source. In my view, the IPO risk for O/A is no longer just a valuation discount, but a substantive challenge to its long-term business model as a high-margin foundation model provider.
In that scenario, even if global Token usage increases 100-fold in the future, that alone would not be enough to justify these foundation model companies being worth one or two trillion dollars.
A nuclear bomb
"Based on this framework, we have achieved high-throughput, low-latency large-scale inference services for DeepSeek V4.1 Flash. Currently, the service throughput of a single cluster is 155 trillion tokens/day, and the overall revenue is $38 per card per hour (at off-peak pricing)"
Source: https://zhuanlan.zhihu.com/p/2095560295930508953
No matter what answer is given, the prediction success rate is 50%. There’s no difference between the model giving a head answer 90% of the time and 50% of the time.
Even if Japan believes that spending money on young people is more appropriate than extending the lives of the elderly by a few years, the decision should absolutely not be to purchase large numbers of American fighter jets, but rather to use the funds to improve the birth rate, build infrastructure, transition to green energy, and support domestic industrial productivity.
I cannot understand what purchasing fighter jets can do for Japan. Japan has in fact completely handed over its military sovereignty to the United States, and buying fighter jets is essentially just giving the United States a blood transfusion.
Yes, a model's technical report should first and foremost include technical details.
A very powerful model, and with multimodal support now, it can be used as a primary model.
All official statements are literal and fragile.
Basically, this means that France, the UK, and the US will use AI in the deployment of conventional weapons.
Just install Pi and start using it directly, no extensions needed for Pi to get the job done.
Only when you genuinely find a feature is missing should you write one yourself or have Pi write an extension or skill for you.
Don't use extensions written by others, after all, those extensions were also vibe-coded.
openrouter’s API has removed parameters specific to certain models, impairing model functionality.
For example, the image generation parameters of GPT-image-2 are largely ineffective.
No microphone input, no speakers, and no included months of Codex membership.
What is the point of this device?
I noticed that the page does not provide uninstall instructions for the sh installation method. Could you please provide them?
My God, what a horrific and evil idea
Victim of LLM hallucinations, poor guy
Even when using OpenRouter in Hong Kong, it is still not possible to connect to region-restricted models like Gemini
This price is high even because of the current shortage of inference cards available to DeepSeek; they claimed in their press release that once the Ascend 950 computing cards are launched in the second half of the year, the price of the Pro version will drop significantly
This is not an official page; deepseek4.hk is not a domain owned by deepseek.
Generating a 4096x4096 image with gemini-3.1-flash-image-preview consumes 2,520 tokens, which is equivalent to $0.151 per image.
Generating a 3840x2160 image with gpt-image-2 consumes 13,342 tokens, which is equivalent to $0.4 per image.
This model is more than twice as expensive as Gemini.
In fact, you need to pay regardless of whether the output includes reasoning tokens or not
Vue provides a computed feature that acts as a buffer layer between the view and the state, so the view and the state are not necessarily strictly bound
Wouldn't it be better to use verified medicine? Even if a patient could see a doctor at any time, the doctor would not prescribe such unverified peptides to the patient
Container ships + modular weapons are too crazy
No matter how large the library, it won't include WhatsApp's API
Oh, bro, you're practically living in 1689
Depending on your time requirements, carving it into a stone tablet is generally a good choice
Mencius, a thinker from over 300 BC, once said: "A gentleman, in his attitude towards animals, having seen them alive, cannot bear to see them die; having heard their cries, cannot bear to eat their flesh. Therefore, a gentleman keeps his distance from the kitchen."
The background of this statement is as follows: King Xuan of Qi once saw a man leading an ox to be slaughtered. Moved by the ox's sorrowful appearance, he was deeply distressed and ordered the butcher to spare the ox. However, the butcher informed him that the ox was intended for sacrificial rites. It must be noted that in ancient times, the two most important affairs of the state were sacrifices and warfare. Sparing the ox would have violated the moral principles of the time. In desperation, King Xuan came up with the idea: "Why not replace this ox with a sheep?" Later, the more he reflected on it, the more absurd he found his own decision. He then sought advice from Mencius, who uttered this statement in response.
Mencius believed that what distinguishes humans from beasts is humanity, meaning that humans treat all things with kindness, but human goodness is limited. Therefore, a gentleman practices kindness by helping those in front of him, unable to extend it to all.
The freshwater used in this osmotic power plant is produced by a wastewater treatment plant, and the high-salinity water used is the wastewater generated by a seawater desalination plant
So, if the freshwater produced by wastewater treatment plants is further processed into usable freshwater to replace the freshwater produced by seawater desalination plants, soon one water treatment plant could replace both the seawater desalination plant and the osmotic power plant, reducing steps. I believe this would greatly improve efficiency
Github Copilot's Edit mode allows you to manually specify the context, and it runs only once each time to write code with diff checking, without entering a agent loop.
The edit mode of GitHub Copilot requires manually providing context files and does not have RAG or other agent tools. I think this mode is much easier to use than the agent mode.