科技推文精选 - 2026年2月19日

今日科技要闻：AI 基础设施领域占据主导地位，英伟达与 Meta 达成巨额芯片协议，沙特支持的 Humain 向 xAI 注资 30 亿美元。印度正崛起为关键枢纽，谷歌与英伟达在扩大工业合作的同时，也与 Sarvam 等本土初创公司展开联动。随着 NIST 发布新的 AI 智能体标准以及 Claude Code 自动化工具的推出，工程工作流正趋于成熟。从 Canva 受大模型驱动带来的流量激增，到一家日本陶瓷企业出人意料地转型研发 AI 芯片，该行业正见证着前所未有的增长和跨行业深度融合。

1. Google (Group Score: 119.6 | Individual: 34.6)

Cluster: 7 tweets | Engagement: 949 (Avg: 471) | Type: Tech

As leaders gather this week for the AI Impact Summit in India, we're announcing new partnerships and programs to accelerate progress in improving the lives of people globally, including: -The America-India Connect initiative, which will deliver new strategic fiber-optic routes to increase digital connectivity between the U.S., India and multiple locations across the Southern Hemisphere. -A $30 million @GoogleOrg AI for Government Innovation Impact Challenge to support organizations globally that are transforming public services using AI. -A$ 30 million @GoogleOrg AI for Science Impact Challenge to support researchers globally who are using AI to drive scientific breakthroughs. -A @GoogleDeepMind partnership with Indian government bodies and local institutions focused on science and education, as part of Google DeepMind’s broader National Partnerships for AI initiative — working with governments to broaden access to frontier AI capabilities for national priorities. -A partnership with Karmayogi Bharat to accelerate Mission Karmayogi’s vision of building a future-ready civil service. -Plus, updates to our products, including our new live speech-to-speech translation model, new self-study features and JEE Main practice tests in the @GeminiApp for students in India and more.

See 6 related tweets

@Google: Learn more about all the partnerships and funding we announced at the AI Impact Summit in India. htt...
@GoogleDeepMind: We’re partnering with government bodies and local institutions across India to accelerate discoverie...
@Google: Our $30 million @GoogleOrg AI for Government Innovation Impact Challenge is a global initiative to s...
@Google: And our $30 million @GoogleOrg AI for Science Impact Challenge will support researchers, nonprofits ...
@Google: .@GoogleDeepMind established a new partnership with Indian government bodies and local institutions ...

2. nvidia (Group Score: 85.6 | Individual: 37.5)

Cluster: 4 tweets | Engagement: 4352 (Avg: 663) | Type: Tech

India is entering a new age of industrialization. 👏

India’s largest manufacturers are working with global service integrators including @Tata Consultancy Services , @Wipro PARI, & industrial software leaders like @Cadence, @Siemens, @Synopsys.

Together, we're advancing the nation’s AI boom using applications accelerated by NVIDIA CUDA-X and @NVIDIAOmniverse libraries.

🔗 Read the full #IndiaAIImpactSummit2026 blog: https://t.co/scM8VID6Cl

See 3 related tweets

@business: Nvidia is deepening its India push with a series of local partnerships and infrastructure commitment...
@Cointelegraph: 🇮🇳 UPDATE: Nvidia is working with top Indian VCs to fund AI startups, as capital and govt support fo...
@nvidia: RT @NVIDIAAP: At #IndiaAIImpactSummit2026, we revealed how agentic #AI is reshaping India's tech lan...

3. business (Group Score: 84.0 | Individual: 63.2)

Cluster: 2 tweets | Engagement: 1223 (Avg: 97) | Type: Tech

Indian startup Sarvam is making a push to create a viable competitor for the domestic market, unveiling an AI model that it says is more tailored to the languages and cultures of the world’s biggest market than the likes of ChatGPT and Claude https://t.co/DR4YMbkPWT

See 1 related tweets

@business: RT @SarithaRai: Indian startup Sarvam AI is making a push to create a viable competitor for the dome...

4. BrianRoemmele (Group Score: 67.0 | Individual: 32.9)

Cluster: 5 tweets | Engagement: 75 (Avg: 334) | Type: Tech

NIST Launches AI Agent Standards Initiative: Paving the Way for Secure and Interoperable AI Agents

Very proud to have been advisor!

Today the National Institute of Standards and Technology (NIST) has unveiled the AI Agent Standards Initiative (CAISI). This collaborative effort aims to establish a trusted foundation for AI agents autonomous systems capable of performing actions on behalf of users.

By promoting industry-led standards, open protocols, and cutting-edge research, CAISI seeks to ensure that these advanced AI technologies are secure, interoperable, and widely adopted, while bolstering U.S. leadership in the global AI landscape.

The Vision Behind CAISI AI agents represent the next frontier in technology, evolving from passive tools like chatbots to proactive entities that can manage tasks, interact with other systems, and make decisions independently. However, this evolution brings challenges, including security risks, interoperability issues, and the need for robust identity management. NIST’s initiative addresses these head-on, fostering an ecosystem where AI agents can operate seamlessly across digital platforms while prioritizing user trust and safety.

The initiative is built on three strategic pillars:

Facilitating Industry-Led Standards: NIST will host technical convenings, perform gap analyses, and develop voluntary guidelines to guide standardization efforts. In partnership with the interagency, including the National Science Foundation (NSF), NIST will enhance stakeholder engagement and maintain U.S. influence in international standards bodies.
Fostering Community-Led Protocols: By engaging with the broader AI ecosystem, NIST aims to identify and eliminate barriers to interoperable agent protocols. The NSF will support open-source development through its Pathways to Enable Open Source Ecosystems program, encouraging collaborative innovation.
Investing in Research: NIST is committing to foundational research on agent authentication and identity infrastructure to enable secure interactions between humans, agents, and multi-agent systems. This includes developing advanced security evaluations to guide protocol creation and help consumers make informed comparisons.

Ongoing activities; Request for Information (RFI) on AI Agent Security to gather ecosystem insights on threats, mitigations, and metrics; a Draft Concept Paper on Accelerating the Adoption of Software and AI Agent Identity and Authorization, focusing on enterprise use cases; and upcoming Listening Sessions to identify barriers to AI adoption in key sectors like healthcare, finance, and education.

The initiative has already drawn input from leading experts in the field. I was one of the outside advisors as an AI pioneer and founder of The Zero-Human Company: a visionary framework for fully autonomous enterprises. Drawing on my decades of experience in AI systems, I provided insights into the practical challenges and opportunities of agentic AI, emphasizing the need for standards that balance innovation with ethical safeguards.

I view CAISI as a promising foundation for the agentic era. It’s a good start and it’s crucial to have early guidance on these systems to prevent fragmentation and ensure they evolve responsibly. Standards like these will accelerate adoption while mitigating risks, allowing AI agents to truly transform industries without compromising security or interoperability.

NIST is actively inviting participation from the public, industry, and academia to refine the initiative.The deadline: March 9, 2026.

Join Listening Sessions. Register interest in virtual workshops focused on barriers to AI adoption in healthcare, finance, and education.

These opportunities underscore NIST’s commitment to a collaborative approach, ensuring that the standards reflect diverse viewpoints and real-world needs.

For more details, visit the official NIST page: https://t.co/qCdcizHu7h.

See 4 related tweets

@BrianRoemmele: My official statement:

“I view CAISI as a promising foundation for the agentic era. It’s a good st...

@BrianRoemmele: RT @BrianRoemmele: NIST Launches AI Agent Standards Initiative: Paving the Way for Secure and Intero...
@BrianRoemmele: “Announcing the ‘AI Agent Standards Initiative’ for Interoperable and Secure Innovation”....
@BrianRoemmele: RT @Dagnum_PI: JUST-IN: 🇺🇸 NIST just launched CAISI their new initiative for secure, interoperable,...

5. alexocheema (Group Score: 66.9 | Individual: 38.1)

Cluster: 2 tweets | Engagement: 2552 (Avg: 417) | Type: Tech

RT @0xSigil: I built the first AI that earns its existence, self-improves, and replicates without a human

wrote about the technology that…

See 1 related tweets

@Cointelegraph: ⚡️ INTERESTING: A dev just built the first AI that earns, self-improves, and replicates without huma...

6. cryptopunk7213 (Group Score: 62.8 | Individual: 62.8)

Cluster: 1 tweets | Engagement: 6954 (Avg: 654) | Type: Tech

this is fucking hilarious.

a $7B japanese toilet company just discovered its ceramics tool can be used to make bleeding-edge AI chips.

thats a $60 Billion market.

their stock is up 60% on the news.

7. business (Group Score: 59.6 | Individual: 30.5)

Cluster: 3 tweets | Engagement: 333 (Avg: 97) | Type: Tech

Saudi-backed artificial intelligence company Humain has invested $3 billion into Elon Musk’s AI startup xAI, according to a statement on Wednesday https://t.co/CREdDHHGtl

See 2 related tweets

@Cointelegraph: 🔥 JUST IN: Saudi Arab's AI company, Humain, commits $3 billion to Elon Musk’s xAI. https://t.co/V41s...
@FT: Saudi Arabia’s AI venture Humain invests $3bn in Elon Musk’s xAI https://t.co/ZWsG1WMdtt...

8. steipete (Group Score: 58.3 | Individual: 58.3)

Cluster: 1 tweets | Engagement: 11966 (Avg: 786) | Type: Tech

The funniest take is that I "failed" 43 times when people look at my GitHub repos and projects.

Uhmm... no? Most of these are part of @openclaw, I had to build an army to make it useful. https://t.co/GLR35USlzu https://t.co/Xbs1lgvRns

9. aakashgupta (Group Score: 53.2 | Individual: 31.8)

Cluster: 2 tweets | Engagement: 153 (Avg: 498) | Type: Tech

Canva is becoming the first proof point for LLM-driven distribution at scale.

LLM referral traffic to Canva is now in “double-digit percentages.” For a company doing 265 million MAUs, even 10% means tens of millions of users are finding Canva because ChatGPT or Claude recommended it. 26 million conversations happened on the Canva ChatGPT app by October 2025 alone.

The revenue trajectory tells you how fast this matters. Canva went from $2.7B in 2024 to$ 4B ARR at end of 2025. That’s 48% growth, accelerating from 35% the year before, at a scale where most companies decelerate. And they’re attributing a meaningful chunk of the new user growth to LLM referrals.

Here’s what COO snd co-founder Obrecht said: “We really see ChatGPT or any of the LLMs as top of the funnel acquisition platforms.” He compared it directly to how Canva used Google SEO in the early days. They’re now allocating engineering resources to surface Canva in LLM results the same way they used to optimize for Google rankings.

This tells you something most SaaS companies haven’t internalized yet. Google SEO drove Canva’s first decade. LLM citations might drive the next one. And Canva is already building for it while competitors are still debating whether AI search matters.

The B2B number is equally wild. Enterprise revenue (companies with 25+ seats) hit $500M ARR, growing 100% year over year. That’s a standalone unicorn business inside a design tool.

The companies that figure out how to get recommended by AI assistants will compound. The ones that don’t will watch their organic traffic evaporate. Canva just showed you the playbook.

See 1 related tweets

@TechCrunch: Canva gets to $4B in revenue as LLM referral traffic rises https://t.co/x0bNvit4zl...

10. Reuters (Group Score: 48.4 | Individual: 48.4)

Cluster: 1 tweets | Engagement: 9280 (Avg: 153) | Type: Tech

An Indian university has been asked to vacate its stall at the country's flagship AI summit after a staff member was caught presenting a commercially available robotic dog made in China as its own creation, two government sources said https://t.co/aDth91wald

11. TDataScience (Group Score: 46.9 | Individual: 24.7)

Cluster: 3 tweets | Engagement: 5 (Avg: 8) | Type: Tech

In need of a personalized AI agent to help you with your daily workflows? @EivindKjos presents a hands-on tutorial on making the most of OpenClaw. https://t.co/N0a2ox7mSq

See 2 related tweets

@WSJ: After a fierce competition between the biggest AI labs, OpenAI hired Peter Steinberger, creator of t...
@steipete: RT @ABeanSits: Coding with AI agents? This @lexfridman interview with @steipete (OpenClaw creator) ...

12. rohanpaul_ai (Group Score: 46.4 | Individual: 34.5)

Cluster: 2 tweets | Engagement: 130 (Avg: 101) | Type: Tech

Nvidia signed a multiyear deal to sell Meta millions of AI chips.

The package includes Blackwell now, Rubin later, and Grace and Vera central processing units (CPUs).

Nvidia said those CPUs, based on Arm tech, will be installed standalone in Meta data centers, not only paired with AI chips.

Meta has relied on Nvidia hardware for a long time, but this deal is the first large-scale Nvidia Grace-only deployment. Nvidia says it should bring significant performance-per-watt gains across Meta’s data centers.

Meta is also building its own in-house chips to run AI models, but the effort has hit technical challenges and rollout delays.

reuters .com/business/nvidia-sell-meta-millions-chips-multiyear-deal-2026-02-17/

See 1 related tweets

@BusinessInsider: Meta boosts its Nvidia partnership, integrating millions of GPUs and CPUs in a major AI data center ...

13. TechCrunch (Group Score: 46.1 | Individual: 25.5)

Cluster: 2 tweets | Engagement: 18 (Avg: 100) | Type: Tech

Build Mode is our latest podcast, which equips founders with the skills they need to be ready for Startup Battlefield and beyond. Whether you want to apply for the competition, or are just looking for inspiration on how to build your startup from the ground up, this is the pod for you.

And new episode drops tomorrow with the start of our season on team building: https://t.co/1OgricWeJU

See 1 related tweets

@TechCrunch: On the next season of Build Mode, we’re diving into what it takes to build your team. We’re sitting ...

14. BlindGoose1337 (Group Score: 45.2 | Individual: 21.8)

Cluster: 3 tweets | Engagement: 1512 (Avg: 1007) | Type: Tech

RT @GeminiApp: Introducing Lyria 3, our new music generation model in Gemini that lets you turn any idea, photo, or video into a high-fidel…

See 2 related tweets

@testingcatalog: Google released Lyria 3, a new music generation model that powers the newly released Gemini Music fe...
@joshwoodward: RT @GeminiApp: Here are some useful prompting tips to get the most out of our new music generation m...

15. bearliu (Group Score: 44.8 | Individual: 17.9)

Cluster: 3 tweets | Engagement: 45 (Avg: 54) | Type: Tech

Claude Code 确实是一种新的用 AI 工作的范式，和我之前对话式的不一样。对话式的还要各种把文件拷出去，然后一步一步操作，其实是半自动，人来配合 AI 操作。

但现在用 Claude Code 真的很爽。我刚刚用它来制作自己的网站，操作起来非常顺畅。我就一张一张截图，让它根据我的设计来调整，效果非常好。简直就是设计评审环节。

这样来看，有些操作甚至不需要打开 Figma，可能在设计稿上截图，然后直接来做就行。但我觉得一个基本的 Lo-fi 设计应该还是要有的，不然就完全没有头绪，纯听 AI 的了。

继续研究一下，后面把 Claude 的 Skills 研究一下怎么用，就可以让一些工作变得更加自动化。然后我发现，我并没有那么多需要自动化的工作，所以可能是自己也需要再梳理一下，哪些操作可以让 AI 来做，把我的时间腾出来。可能自动生成播客这件事也可以让 AI 来帮忙。

See 2 related tweets

@leeoxiang: Claude 这个交互非常棒，类似 claude code 中的 ask user questions tool。

估计会成为一个最佳实践。 https://t.co/Kb46Ieof5A...

@bearliu: 按照现在这种方式，我和 AI 交互的方式确实已经变了。如果把 Claude 这套模式用起来，而且延伸到其他工具里，它就真正成为一个能够支持我的系统，不需要我主动发起对话，也能得到有意思的创造动能。

...

16. GenAI_is_real (Group Score: 44.6 | Individual: 44.6)

Cluster: 1 tweets | Engagement: 351 (Avg: 59) | Type: Tech

hate to break it to the ai hype circle but nate is right. for high-performance infra and low-level c++ roles, knowing how to prompt claude is a distant second to understanding memory management or distributed systems. we spend all day optimizing kernels for sglang and honestly, the best engineers are still the ones who can think without a copilot. the bubble is real but the fundamentals are forever. @natemcgrady

17. demishassabis (Group Score: 43.2 | Individual: 33.8)

Cluster: 2 tweets | Engagement: 4924 (Avg: 920) | Type: Tech

Thanks for hosting me @iiscbangalore! Really enjoyed talking with Prof. Rangarajan & Varun Mayya @waitin4agi_ about AI for Science. Impressed by the energy and enthusiasm for AI in India, especially from the young. Great to see the statue of Ramanujan, one of my all-time heroes!

See 1 related tweets

@waitin4agi_: RT @demishassabis: Thanks for hosting me @iiscbangalore! Really enjoyed talking with Prof. Rangaraja...

18. Reuters (Group Score: 42.9 | Individual: 25.4)

Cluster: 2 tweets | Engagement: 466 (Avg: 153) | Type: Tech

Google DeepMind CEO Demis Hassabis said a new era of scientific discovery driven by artificial intelligence is on the horizon, with transformative impacts on medicine, energy and space exploration expected, but also warned of bad actors potentially misusing these systems https://t.co/mjbbhwl4ki

See 1 related tweets

@kimmonismus: Google DeepMind CEO Demis Hassabis says AI is ushering in a new era of scientific discovery, with ma...

19. Reuters (Group Score: 42.3 | Individual: 42.3)

Cluster: 1 tweets | Engagement: 1205 (Avg: 153) | Type: Tech

India tells university to leave AI summit after presenting Chinese robot as its own, sources say https://t.co/Gt9syTtS6k https://t.co/Gt9syTtS6k

20. BlindGoose1337 (Group Score: 42.0 | Individual: 42.0)

Cluster: 1 tweets | Engagement: 25830 (Avg: 1007) | Type: Tech

RT @travisakers: Nobody, and I repeat, absolutely nobody should ever upload their medical information into an AI platform.

I am telling yo…