Odaily星球日报讯 曾任以太坊开发商 ConsenSys 全球数字资产架构师的 Ejaaz Ahamadeen 于 X 发文表示：“DeepSeek 只是加速了所有时间表。我们之前接受了‘事实’，这些产品的构建将花费 1000 亿美元，它将是闭源的，意味着形成垄断。 然后 DeepSeek 以极低的成本超越了所有领先模型，并且开源了这项技术，现在任何 AI 构建者都可以访问这款最强大的模型。 为什么没有人谈论我们将如何看到‘更好的代理’？胖应用理论变得更加真实。这对开源 AI 来说很棒。”

Odaily星球日报讯 曾任以太坊开发商 ConsenSys 全球数字资产架构师的 Ejaaz Ahamadeen 于 X 发文表示：“DeepSeek 只是加速了所有时间表。我们之前接受了‘事实’，这些产品的构建将花费 1000 亿美元，它将是闭源的，意味着形成垄断。然后 DeepSeek 以极低的成本超越了所有领先模型，并且开源了这项技术，现在任何 AI 构建者都可以访问这款最强大的模型。为什么没有人谈论我们将如何看到‘更好的代理’？胖应用理论变得更加真实。这对开源 AI 来说很棒。”

开发者：DeepSeek以极低成本超越所有领先模型并开源，加速了时间表

Odaily reported that Ejaaz Ahamadeen, former global digital asset architect at <ConsenSys>, wrote in a post on X that "DeepSeek has only accelerated all timelines. We had previously accepted the 'fact' that the construction of these products would cost $100 billion, and they would be closed-source, meaning they would form a monopoly. Then DeepSeek surpassed all leading models at an extremely low cost and open-sourced this technology, and now any AI builder can access this most powerful model. Why isn't anyone talking about how we will see 'better agents'? The fat application theory is becoming more real. This is great for open-source AI."

Odaily reported that Ejaaz Ahamadeen, former global digital asset architect at <ConsenSys>, published an article stating: "DeepSeek has only accelerated all timelines. We had previously accepted the 'fact' that the construction of these products would cost $100 billion, and they would be closed-source, meaning they would form a monopoly. Then DeepSeek surpassed all leading models at an extremely low cost and open-sourced this technology, and now any AI builder can access this most powerful model. Why is no one talking about how we will see 'better proxies'? The fat application theory is becoming more real. This is great for open-source AI."

Developer: DeepSeek surpasses all leading models at a very low cost and is open source, accelerating the timeline

Here is the English translation:

Odaily report: Recently, DeepSeek has sparked heated discussions, and the founder of Black Myth: Wukong commented on social media that it could be a national-level technological achievement. The full text is as follows:

"Let's seriously promote DeepSeek before the end of the year. I've been using V3 for a month, but R1 for only 5 days, so I have to post this Weibo.

First, let me make a bold statement: DeepSeek may be a national-level technological achievement.

To explain how amazing this achievement is, let me give you an example: If an AI large model can achieve any of the following, it would be an incredible breakthrough:

1. Powerful. Comparable to GPT-3's reasoning ability, unparalleled for now.
2. Inexpensive. The number of parameters is small, and the training and usage costs are an order of magnitude lower.
3. Open-source. Anyone can download and deploy it, with detailed papers explaining the training steps and tricks, and even providing a mini-model that can run on a mobile phone.
4. Free. The official service is completely free, and anyone can use it anytime, anywhere.
5. Networked. The only reasoning model that currently supports online search (GPT-3 does not support this yet).
6. Homegrown. DeepSeek is developed by a small, young Chinese company, with a local team that has no overseas experience or senior industry experience.

DeepSeek has achieved all six of these. It's truly amazing, almost unbelievable.

If this doesn't deserve a cheer, then what does?

So, what should you do now that you've read this Weibo?

1. Go directly to the website and start using it, there's also a mobile app.
2. Use it intensively, try to replace traditional search with it. Imagine you know the top experts in every field, and they are all your good friends online. Ask them anything you might need to "consult someone" about - whether it's building a complex function, figuring out which swimming pool is best nearby, checking for legal traps in a contract, helping refine song lyrics, or verifying the effectiveness of a health product for your parents.
3. See how others are using it, try other large models, understand AI's strengths and weaknesses, how to fine-tune it, and then continue to unlock and iterate your own usage and more tools.

I hope DeepSeek R1 will demystify the current most advanced AI for you, and gradually make AI become like water and electricity in your life.

So lucky! So happy! This stunning breakthrough comes from a pure Chinese company.

Another solid step towards equality of knowledge and information."

The founder of Black Myth Wukong commented on DeepSeek: It may be a national technological achievement

The DeepSeek-R1 uses a pure reinforcement learning approach and has achieved performance on par with top models such as GPT-4o and Claude Sonnet 3.5.
Tencent Technology's "AI Future Compass" special author: Hao Boyang
In less than a month, DeepSeek has once again shaken the global AI community.
Last December, DeepSeek's DeepSeek-V3 caused a huge sensation in the global AI field. It achieved performance comparable to GPT-4o and Claude Sonnet 3...

DeepSeek's new model reveals why it can shake the global AI circle?

[Important News from Last Night and This Morning (January 26 - January 27)]

DeepSeek tops the free app chart in Apple's China App Store
According to the free app chart on Apple's App Store in China, Deepseek has become the number one app in China.

DeepSeek surpasses ChatGPT, topping the free app download chart on Apple's US App Store
The Deepseek app developed by the domestic large model company Hangzhou Shendu Qiusuo topped the free app download chart on the Apple App Store in both China and the US on January 27, surpassing ChatGPT in the US download chart.

Vitalik calls for accelerating the elimination of Groth16 trusted setup
Jupiter announces the destruction of 3 billion JUP tokens
Jupiter: 50% of protocol fees will be used to buy back JUP
Solana ecosystem decentralized exchange Jupiter stated on the X platform: "From now on, 50% of Jupiter protocol fees will be used to buy back JUP."

Pudgy Penguins CEO: Abstract mainnet will launch on Monday
Michael Saylor releases Bitcoin Tracker-related information again
CZ reveals that he first bought Bitcoin in early 2014, when the price was around $600
Analyst: Estimated $50 million will be used to buy back JUP, equivalent to 2.4% of JUP's current market value

Pump.fun sets a single-day fee income record of $15.38 million
According to The Block data, the Meme coin platform Pump.fun reached a fee of $15.38 million on January 24, setting a single-day high record, partly driven by the hype of the Meme coin Vine. The platform has processed $4 billion in transactions in the past two weeks and has generated nearly $250 million in SOL fees since its launch over a year ago.

ZachXBT: Suspected hacker stole about $29 million on the Sui network last month
Viewpoint: Deepseek disrupts the traditional perception of the AI industry, and the market faces a reshuffle in the future

Data: Tokens such as SUI, OP, and TRIBL will face large-scale unlocking next week, with SUI unlocking about $265 million
Token Unlocks data shows that tokens such as SUI, OP, TRIBL, and ZETA will face large-scale unlocking this week:
Sui (SUI) will unlock about 64.19 million tokens at 8 am Beijing time on February 1, accounting for 2.13% of the current circulating supply, valued at about $265 million;
Optimism (OP) will unlock about 31.34 million tokens at 8 am Beijing time on January 31, accounting for 2.32% of the current circulating supply, valued at about $51.4 million;
Tribal Token (TRIBL) will unlock about 22.07 million tokens at 8 am Beijing time on January 29, accounting for 14.03% of the current circulating supply, valued at about $39.7 million;
ZetaChain (ZETA) will unlock...