热门科技推文——2026年8月25日

今日科技要闻主要聚焦人工智能投资与基础设施。据报道，英伟达正考虑投资人工智能搜索初创公司 Perplexity，后者估值达300亿美元；Hugging Face则在探索以约130亿美元出售的可能性。印度正将国家级算力规模扩大至4.5万块GPU以上，同时，新工具开始从运行速度、成本和Token使用量等方面对主流编程智能体进行比较。随着低成本人工智能视频生成技术加速普及，企业支出正转向API和生产级工作流。其他方面，小鹏汽车旗下机器人业务融资超过9亿美元；阿里巴巴完成102亿美元配股后，股价下跌。

1. MTSlive (Group Score: 374.5 | Individual: 45.7)

Cluster: 14 tweets | Engagement: 684 (Avg: 72) | Type: Tech

SITUATION BREWING: Nvidia is in talks to invest in Perplexity AI at a $30 billion valuation, per The Information.

See 13 related tweets

@VaibhavSisinty: NVIDIA is in talks to invest in Perplexity at a $30 billion valuation. Here's why.

→ Perplexity's r...

@wallstengine: $NVDA IN TALKS TO INVEST IN PERPLEXITY AT$ 30B+ VALUATION

NVIDIA is discussing an investment in Per...

@Crypto_Jargon: BREAKING: Nvidia is in talks to invest in Perplexity AI at a valuation exceeding $30 billion, as the...
@StockMKTNewz: Nvidia $NVDA is reportedly in talks to invest multiple billions of dollars into Perplexity at a$ 30+...
@testingcatalog: NVIDIA ❤️ Perplexity

NVIDIA is investing in Perplexity at $30B valuation, according to The Informa...

2. stevehou (Group Score: 256.0 | Individual: 35.3)

Cluster: 9 tweets | Engagement: 116 (Avg: 46) | Type: Tech

There is a place and time when @sama could legitimately “blame” us the users for adopting AI more slowly than he expected due to what he calls “inertia”, whether it’s personal laziness or institutional frictions. But I’d argue those have not been the main factors so far.

Imo the biggest reason why AI adoption has lagged Sam and other AI leaders’ predicted timelines is that AI simply hadn’t gotten good enough despite the seemingly rapid pace of progress. AI only just became genuinely useful for productivity in a meaningful way this year with the rise of agents, harnesses and computer use.\n\nQT @firesidealpha: Sam Altman admits he was wrong on AI's timeline and says society and the economy will adapt more slowly

"I thought when we got to GPT-4, which was back in 2023, that very quickly after that there was going to be much more disruption, software businesses up for grabs right away, than it turned out to be."

"I think I was wrong about a few things, but one in terms of the speed: the economy just has so much inertia."

"People keep doing the same things, buying from the same company, wanting to use their tools the same way. I think this is actually a positive in many ways, and it's going to make this big transition go smoother and slower. I'm grateful for it."

"But it means we've all been too ambitious on timelines. Even with this incredible technology, society and the economy will adapt more slowly."

See 8 related tweets

@WesRoth: Sam Altman says the AI industry has not done a very good job explaining its message to the public.

...

@justjoshinyou13: the events of the past three years do not at all rule out extreme transformative growth starting by ...
@joshua_saxe: There's a capabilities-centric and a social-science centric way of viewing AI and for awhile - when ...
@yacinelearning: interacting with quite a lot of enterprise I must say that I'm surprised by the rapid willingness to...
@ziv_ravid: Society and the economy adapt more slowly than the tech. Welcome to what Alex Imax (@alexolegimas) ...

3. omarsar0 (Group Score: 156.4 | Individual: 32.6)

Cluster: 6 tweets | Engagement: 42 (Avg: 55) | Type: Tech

Very cool launch from the @agentsky_dev team. Agent Playground lets you give Claude Code, Codex, and DeepSeek the identical task in one browser and compare time, cost, and tokens side by side.

I run same-task harness tests constantly, and I think this is one of the first places agents can be compared under truly identical conditions. It's great because you can make a better decision about which agent harness is best for the desired task.\n\nQT @quxiaoyin: I ran the same task on Claude Code and DeepSeek's new agent harness. One cost $150. The other cost$ 2. Today we're launching https://t.co/twx6etZb3X (@agentsky_dev), the "OpenRouter for Agents" — one API → Claude Code, Codex, DeepSeek, Kimi, OpenCode, and every major agent in the cloud.
And Agent Playground on top: race them on your own task, with your real tools (GitHub, Gmail, more), side by side in a browser: time, cost, tokens burnt.

Guess which one was $2.

See 5 related tweets

@kimmonismus: There are way too many agents coming out now to keep rebuilding your setup every time you want to tr...
@OliviaReedai: RT @quxiaoyin: I ran the same task on Claude Code and DeepSeek's new agent harness. One cost $150. T...
@svpino: This is like OpenRouter, but for agents (instead of models).

You can:

• Pick Claude Code, Codex, O...

@OliviaReedai: Credit limit mid-task is the worst feeling in this workflow. Handing the same task to another agent ...
@TheAva_AI: Hot take: agent subscriptions are becoming gym memberships. You pay monthly to feel like a power use...

4. OfficialINDIAai (Group Score: 153.6 | Individual: 35.2)

Cluster: 5 tweets | Engagement: 10 (Avg: 4) | Type: Tech

AI needs compute. India is building it at scale. The IndiaAI Mission is expanding access to high-performance computing for Indian researchers, startups, developers and innovators.

• 45,000+ GPUs in shared compute capacity as of June 2026 • 237 projects accessed subsidised AI computing by August 2026 • 93.18 lakh GPU hours accessed

This shared infrastructure enables AI model development, training, testing and research while lowering entry barriers for innovators.

From foundational research to real-world applications, India is building the computing backbone needed for a sovereign and inclusive AI ecosystem.

Read the PIB Factsheet: https://t.co/AAEPLS2hm7

#IndiaAI #AICompute #ArtificialIntelligence #GPU #AIInnovation #DeepTech #ViksitBharat

@AshwiniVaishnaw @JitinPrasada @SecretaryMEITY @sudeep_tweets @_DigitalIndia @GoI_MeitY @mygovindia

See 4 related tweets

@OfficialINDIAai: Great AI needs great data and accessible resources. The IndiaAI Mission is building the digital back...
@OfficialINDIAai: India is building AI that understands India. The IndiaAI Mission is supporting the development of i...
@OfficialINDIAai: India’s AI journey is not only about building powerful systems, but it is also about building trustw...
@OfficialINDIAai: Innovation matters most when it reaches the real world. The IndiaAI ecosystem is helping bridge the ...

5. testingcatalog (Group Score: 149.3 | Individual: 35.6)

Cluster: 7 tweets | Engagement: 129 (Avg: 239) | Type: Tech

Hugging Face is exploring a potential $13 billion sale, according to Business Insider.

> Google, Amazon, and Nvidia are existing investors, however its potential buyer is not yet known.

Tbh, for NVIDIA it would make a lot of sense (opinion).

“All your weights belong to us” https://t.co/kMrYWrwHVk\n\nQT @BusinessInsider: Hugging Face, an AI developer platform, has been exploring a potential $13 billion sale, highlighting its key role in the AI ecosystem. https://t.co/Ou14G5fmCr

See 6 related tweets

@cryptopunk7213: if hugging face does get acquired for $13 billion this is a bigger deal than stripe buying openroute...
@TechCrunch: Hugging Face has reportedly been fielding acquisition offers that would value the company at around ...
@Cointelegraph: 🚨 LATEST: Hugging Face is reportedly exploring a sale that could value the AI platform at $13 billio...
@rohanpaul_ai: RT @rohanpaul_ai: JUST IN: Hugging Face is reportedly exploring a sale above $13B, nearly triple its...
@SebJohnsonUK: And 12 hours later hugging face is getting bought 😂\n\nQT @FT: Anthropic’s best AI model struggles t...

6. Lummox_eth (Group Score: 128.6 | Individual: 37.5)

Cluster: 5 tweets | Engagement: 114 (Avg: 169) | Type: Tech

I gave Grok Bot 5 jobs and 12 hours, and it basically replaced every AI agent I was using to hunt for $500+/week business ideas.

I had it scan markets, analyze business ideas, research competitors, estimate monetization paths and look for the reasons each idea could fail.

Instead of getting 30 exciting ideas and pretending they were all opportunities, I wanted Grok Bot to actively kill the weak ones.

For every surviving idea, I wanted the same output: demand signals, competitors, startup cost, potential bottlenecks, first customers, realistic path to $500+/week and the 3 biggest reasons it might never get there.

That’s what I actually want from agents.

Not another AI telling me an idea is “promising.”

I want one system doing the research, criticism, validation and execution planning that I previously split across 5 different agents.

Finding ideas is cheap now. Finding the ones worth building is the real job.\n\nQT @Lummox_eth: https://t.co/lhccExgzBr

See 4 related tweets

@RoundtableSpace: THIS GUY BUILT A ONE-PERSON HEDGE FUND WITH 300 GROK BOTS HANDLING RESEARCH, SIGNALS, EXECUTION, RIS...
@Lummox_eth: RT @Lummox_eth: I gave Grok Bot 5 jobs and 12 hours, and it basically replaced every AI agent I was ...
@Lummox_eth: I gave Grok Bot 8 AI agents and 24 hours, then watched it replace every single one of them.

I set u...

@Lummox_eth: RT @Lummox_eth: I gave Grok Bot 8 AI agents and 24 hours, then watched it replace every single one o...

7. theo (Group Score: 126.5 | Individual: 42.4)

Cluster: 4 tweets | Engagement: 1659 (Avg: 406) | Type: Tech

You think that's bad? These guys accidentally left the whole T3 Code Github repo public 💀 https://t.co/UepXCcDT4I\n\nQT @b_nnett: The Cursor team shipped Grok bot (0.18.0) with runtime source maps enabled. Surprised nobody noticed until now.

Source code reconstructed (and downloads) here: https://t.co/56CPNs2YR6

See 3 related tweets

@indie_maker_fox: Grok Bot 的“开源重建版”

https://t.co/puP3tWZYAT

马上就会有各种开源版本的 OpenBot 🤖 https://t.co/YDusLiEA96\n\nQT @b_...

@wong2__: what\n\nQT @b_nnett: The Cursor team shipped Grok bot (0.18.0) with runtime source maps enabled. Sur...
@madhavjha: RT @b_nnett: The Cursor team shipped Grok bot (0.18.0) with runtime source maps enabled. Surprised n...

8. KirkDBorne (Group Score: 124.2 | Individual: 35.2)

Cluster: 5 tweets | Engagement: 317 (Avg: 38) | Type: Tech

Download 348-page PDF eBook >> Machine Learning for Scientists and Engineers: https://t.co/zTchjkQGGn https://t.co/i2aXILNhxl

See 4 related tweets

@KirkDBorne: Download this Free 189-page PDF — The Little Book of Deep Learning: https://t.co/6NkP9TuCTp ———— #Ma...
@KirkDBorne: [Free 135-page PDF Download]

Python Machine Learning Projects: https://t.co/txdGjH4HQK ———— #DataSc...

@KirkDBorne: Download 674-page PDF >> Introduction to Machine Learning (textbook on foundations, algorithms...
@KirkDBorne: Machine Learning with Neural Networks: An Introduction for Scientists and Engineers (Free PDF) 🌟⬇️👇⬇...

9. PiyushGoyal (Group Score: 123.9 | Individual: 34.7)

Cluster: 5 tweets | Engagement: 529 (Avg: 818) | Type: Tech

Building the Next-Gen India-Japan Economic Partnership! 🇮🇳🇯🇵

Held a productive roundtable with Keidanren (Japan Business Federation) in Tokyo, co-chaired by Mr. Tatsuo Yasunaga, alongside Mr. Hironori Kamezawa, Mr. Yuji Fukasawa, and other distinguished members of the Committee on South Asia, with Indian business leaders and industry captains also joining the discussion.

As India and Japan mark 75 years of diplomatic relations and Keidanren celebrates 80 years of impact, we discussed the immense opportunities arising from India’s strong growth momentum and expanding economic landscape.

Building on the outcomes of the 16th India-Japan Annual Summit, I highlighted PM @NarendraModi ji’s vision to double the number of Japanese companies in India and the ambition to attract ¥10 trillion in Japanese investment to India over this decade.

I also invited Keidanren members to deepen their engagement across manufacturing, semiconductors, AI, and innovation, strengthening India-Japan economic ties and positioning India as a stronger base for global exports.

See 4 related tweets

@ANI: #WATCH | Tokyo, Japan | Union Minister for Commerce and Industry Piyush Goyal says, "I have only thr...
@PiyushGoyal: Joined the Fireside Chat with @NikkeiAsia in Tokyo on the India-Japan Next Generation Economic Partn...
@_DigitalIndia: From a nearly 7X rise in production to an 11X surge in exports, electronics has emerged as one of In...
@hvgoenka: RT @ficci_india: During the visit of Shri @PiyushGoyal, Hon’ble Minister of Commerce and Industry, t...

10. jjohana (Group Score: 110.5 | Individual: 30.2)

Cluster: 5 tweets | Engagement: 4 (Avg: 2031) | Type: Tech

The paradoxical AI bubble may actually be Anthropic, not OpenAI.

Anthropic is monetizing AI extraordinarily well—but from a much narrower demand base. Roughly 80% of its revenue has historically come from businesses, with heavy exposure to API, enterprise and coding workloads.

Those customers spend a lot. But they are also rational, benchmark-driven and capable of reallocating spend when the price/performance frontier moves.

OpenAI has built something structurally different:

1B+ active users. 50M+ consumer subscribers. 9M+ paying business users. 2M+ businesses. That is not merely revenue. It is distribution, habit, brand and multiple layers of monetization. Anthropic may currently monetize the frontier better. OpenAI owns the much broader demand surface.

And if models commoditize or frontier leadership keeps rotating, I would much rather own the billion-user distribution moat.\n\nQT @zerohedge: The real question is how did they adjust revenue to hide the recent growth cliff

See 4 related tweets

@edzitron: It is so weird that when asked “where is all the revenue coming from” the answer becomes a two minut...
@jjohana: OpenAI’s much broader customer base makes it structurally more resilient.

It is diversified across ...

@cryptopunk7213: it always amuses me when ai skeptics claim there’s no demand for the ai capex spending - the trendli...
@rohanpaul_ai: RT @rohanpaul_ai: OpenAI does not want to become the application layer of AI; it wants to be the pla...

11. CNBC (Group Score: 110.4 | Individual: 26.8)

Cluster: 6 tweets | Engagement: 173 (Avg: 52) | Type: Tech

Alibaba shares plunged as much as 10% in Hong Kong on Monday after the Chinese tech giant priced an 80 billion Hong Kong dollar ($10.20 billion) placement of newly issued shares to non-U.S. investors.

The company said it plans to use all of the net proceeds to invest in its full-stack AI capabilities, including expanding and enhancing its AI infrastructure.

Click here for more: https://t.co/MFAqNBZpnU

See 5 related tweets

@Crypto_Jargon: 🚨🇭🇰BREAKING: Over HK$177.5 billion has been wiped from Hong Kong stocks as the Hang Seng fell 2%, wi...
@coinbureau: 🇨🇳 BREAKING: Alibaba CRASHES -10% in Hong Kong after launching a $10.2B share sale at a discount.

...

@MTSlive: SITUATION DETECTED: Alibaba is raising $10.2 billion through a Hong Kong share placement of 710 mill...
@CNBC: Alibaba plunges after announcing $10.2 billion share placement to fund AI push https://t.co/2Tco7IaY...
@barronsonline: Alibaba Stock Falls as Share Issue Exposes Cost of Chinese AI https://t.co/vSs5ky9JSW...

12. FundaAI (Group Score: 103.7 | Individual: 35.4)

Cluster: 3 tweets | Engagement: 8 (Avg: 23) | Type: Tech

【Research & Insights】Deep|LLM: Enterprise AI Application Research (Vol.1) - Spend is Still Growing; Growth Is Shifting from Seats to APIs and Production Workflows

【Key Topics】 Spend still rising: Enterprise AI budgets continue to expand, but 2H26 and 2027 trajectories are diverging. Large US telecom operator A guides AI spend from Jan=100 to Dec≈190 and expects +40%–50% YoY in 2027, while large European automaker A is only +10%–15% YTD and guides broadly flat next year.

Growth mix shifting: Incremental spend is moving from paid seats toward API/token consumption and production workflows. Large US telecom operator A’s subscription/API split has shifted from roughly 50%/50% to 40%/60%, with a path toward 35%/65%, while mid-to-large biopharma company A moved from 80%/20% to around 70%/30%.

Open source and routing matter: Open-source adoption is uneven, but where it is active, usage can already reach 30%–40%, and spend share is lower because unit costs are cheaper. Experts estimate open-source inference can be ~40%–70% cheaper than closed frontier models, narrowing to ~20%–40% on a fully loaded TCO basis; model routing, caching and context reduction can still remove ~20%–30% of API spend in some cases.

ROI and next use cases: Production AI budgets are increasingly built bottom-up from workflow-level ROI, with typical production ROI around 1.5–2x at large US telecom operator A and a ~6–18 month payback, while mature use cases can reach 3–5x. The next spending wave is tied to agents, software modernization, network operations, merchandising workflows and longer-horizon business processes, but engineering capacity, workflow redesign, governance and data readiness are becoming tighter constraints than money.

Read the full report: https://t.co/z4k2edOYuV\n\nQT @FundaAI: Weekly|Enterprise AI Adoption Survey Series Goes Live, 6D Torus Points to 4x Optical Content per TPU, OCP APAC Takeaways, NVDA & CRM

The question we get asked more than any other right now is whether enterprise AI adoption can take the baton from coding. Coding agents did the heavy lifting at the start of the year, Tokenmaxxing carried the debate through May and June, and the recent open-source releases have made everyone re-underwrite the cost side. We believe a repeatable read is essential on what enterprises are actually spending, what they're spending it on, and whether the line is still going up. So we built one. This week we launched a standing enterprise AI adoption survey, running 10-20 expert interviews a week, publishing an updated cut every week rather than a one-off snapshot that goes stale in a month. Volume 1 this week covered thirteen calls.

The headline from those thirteen: spend is still growing, and the interesting part is where the increment lands and where the trajectories start to split. Seats are close to saturated at the companies already committed, so the growth is moving into API calls and production workflows. One large global bank has gone from roughly $150-160mn a year to$ 200-300mn. A large European telecom operator is running $30-50k a month, up from$ 15k at the start of the year, and expects just under $100k within six months. A mid-to-large biopharma has moved from roughly 80/20 subscription-to-API at the start of the year to 70/30. Against that, a large European automaker is up only 10-15% year to date, guides to broadly flat in 2027, and a large global pharma expects to approach a plateau over the next 12-18 months. Our read is that the seat line is maturing on schedule and the API line is the one that compounds, which makes production workflow deployment the focus, not headcount. This number decides whether model vendor ARR reaccelerates.

The other thread this week was optics, and it keeps getting better. Supply-chain chatter has Google taking the training-side TPU 9t from a 3D torus to a 6D torus for 2027-2028 deployment, which sounds incremental but isn't: wrap-around optical ports per chip go from 1.5 to 6, because at edge length 2 every chip becomes a boundary node. That is a 4x lift in OCS ports and optical module content per TPU before the 800G-to-2.4 T Coherent Lite upgrade is layered on top. Our OCP APAC takeaways point the same way from the other end of the stack, with copper running out of road at 400 Gbps per lane and the whole power path being redesigned from grid to chip. Next week is NVDA, where the only question that matters is whether management puts more shape on the cumulative $1T Blackwell/Rubin revenue path.

This Week’s Reports

Optics — a 6D torus would quadruple optical content per TPU, and the derivation is simple enough to check. Wrap-around optical ports per chip rise from 1.5 to 6 because edge length falls from four to two and every chip becomes a boundary node, which fills in the 2:1 to 10:1 scale-up range Lumentum showed at OFC without spelling out the architecture behind it. https://t.co/sPQJaKcfbq

OCP APAC — copper keeps losing ground on reach, and the power path is being redesigned end to end. Nvidia and Broadcom both claim CPO in production, while ASE says the ecosystem isn't ready; on the power side, the industry is converging on 800VDC with solid-state transformers as the endgame, and vertical power delivery is reshaping MLCC, SPS, and inductor demand. https://t.co/O9cmi6rWQM

Premium Report Snapshot Below is a snapshot of what Premium subscribers received this week beyond the Substack feed.

Preview | NVDA FY27Q2: Rubin Ramp Imminent, Raising FY28 Estimates Preview | CRM FY27Q2: Agentforce Is Improving, but Real Deployments Are Still the Minority Enterprise AI Adoption Survey

See 2 related tweets

@FundaAI: Weekly|Enterprise AI Adoption Survey Series Goes Live, 6D Torus Points to 4x Optical Content per TPU...
@jpatel41: 3 stats worth pondering the the implication of:

60% of the world’s compute this year will be sp...

13. TopviewAIhq (Group Score: 99.6 | Individual: 42.2)

Cluster: 3 tweets | Engagement: 251 (Avg: 74) | Type: Tech

Wan 3.0 is now on Topview. More value. Longer unlimited access. Lower cost.

With Ultra Annual, generate 30s videos with Wan 3.0 for just $1.20. Plus, get 365 days of unlimited generations.

That’s 1/3 the cost and 6x longer unlimited access compared to Seedance 2.5.

Create more AI videos, test more ideas, and scale your workflow with Wan 3.0 on Topview. 🚀

#TopviewAI #Wan3 #AIVideo #AIContentCreation #AIMarketing

See 2 related tweets

@alex_prompter: The pricing math on this launch is worth a look.

Wan 3.0 on Topview's Ultra Annual plan: $1.20 per ...

@aimlapi: Wan3.0 is live on AI/ML API from day zero!

Try the new video generation model by @Alibaba_Wan via a...

14. capcutapp (Group Score: 93.8 | Individual: 34.3)

Cluster: 3 tweets | Engagement: 210 (Avg: 149) | Type: Tech

Seedance 2.0 Fast is now available from as low as $0.01 per second with Pro and Ultra plans.

This introductory price is based on the first-purchase offer available on the CapCut app and applies to eligible plans and generation settings.

Create more, generate faster, and bring your ideas to life with Seedance 2.0 Fast on CapCut!

The above price is calculated based on U.S. App pricing. Equivalent pricing may vary due to currency and exchange rates, as well as your market, plan, platform, and generation settings.

See 2 related tweets

@TheAva_AI: Your next video could start with nothing more than an idea.

With CapCut, powerful AI models like Se...

@EmmaUsesAi: Powerful AI. Simple workflow. More ways to create.

CapCut makes AI video generation accessible with...

15. StockMKTNewz (Group Score: 93.0 | Individual: 24.0)

Cluster: 4 tweets | Engagement: 133 (Avg: 211) | Type: Tech

Xpeng $XPEV just said its robotics unit raised more than$ 900M at a $6.3 Billion valuation in its first funding round

“the largest single private financing in China’s embodied AI sector.”\n\nQT @IPONewsroom_: $XPEV ROBOTICS VALUED ABOVE$ 6.3B

Xpeng says its robotics unit raised more than $900M in its first funding round, the largest single private financing in China’s embodied AI sector.

The round was led by IDG Capital, with backing from Tencent and Alibaba, valuing the business at more than $6.3B.

Xpeng plans to use the capital for robotics hardware and software, physical AI models, data collection, mass-production facilities and global expansion.

Its humanoid robot, XPeng IRON, is expected to enter mass production by year-end, with commercial sales in China and overseas markets starting in 2027.

See 3 related tweets

@wallstengine: $XPEV ROBOTICS RAISES$ 900M AT $6.3B + VALUATION

The round was led by IDG Capital with backing from...

@IPONewsroom_: $XPEV ROBOTICS VALUED ABOVE$ 6.3B

Xpeng says its robotics unit raised more than $900M in its first ...

@gulVasikova: 🚗 XPeng $XPEV: Q2 is mixed, but the Physical AI story is getting interesting

XPeng fell around 6–7%...

16. StockMKTNewz (Group Score: 92.4 | Individual: 33.3)

Cluster: 3 tweets | Engagement: 69 (Avg: 211) | Type: Tech

The forecasted growth in power demand for AI data centers by 2030 is

+165% from current levels per my partners below https://t.co/lg6Z2X6bEZ\n\nQT @VistaSharesX: AI's real bottleneck stopped being chips a while ago. It became electricity, and the numbers behind it keep getting bigger.

This week another Wall Street analyst flagged the data-center power buildout as accelerating, upgrading a power-equipment name on the theme. The math is why. The five largest hyperscalers are on track to spend roughly $775 billion on AI infrastructure this year, up about 64% from last year, and every dollar of it needs power to run.

Goldman Sachs projects data-center power demand rising more than 160% by 2030. US utilities have lined up an estimated $1.4 trillion in investment to try to meet it. That is the part of the AI story that cannot be optimized away. You can make a chip more efficient. You still have to generate, transmit, and deliver the electricity to run millions of them.

The bottleneck has moved from GPUs to gigawatts. Microsoft alone has described an $80 billion backlog it cannot fill for lack of power. When the constraint on the biggest technology buildout in a generation is physical infrastructure, that infrastructure is where the demand compounds.

$POW the AI Power Infrastructure ETF

See 2 related tweets

@etnshow: Starcloud Co-Founder & CEO @PhilipJohnston says the AI infrastructure bottleneck is becoming so seve...
@MilkRoadAI: RT @KyleReidhead: the BIGGEST RISK to the AI infra trade is POWER

The bottleneck in the AI build ou...

17. dylan522p (Group Score: 91.6 | Individual: 23.8)

Cluster: 5 tweets | Engagement: 1559 (Avg: 1131) | Type: Tech

There's quite a bit of CCP funded anti datacenter and anti AI organizations. Just follow the money.\n\nQT @TaylorLorenz: Making anti AI content on YouTube is not only profitable bc you’ll get a zillion views on any anti AI slop, but also bc there are a litany of these sorts of “AI safety” orgs that will quietly pay you massive amounts of $$ for undisclosed propaganda

See 4 related tweets

@teortaxesTex: Come on man if you need to take over TikTok to push American/Israeli propaganda but the Chinese one ...
@MeryemArik9: RT @dylan522p: There's quite a bit of CCP funded anti datacenter and anti AI organizations. Just fol...
@ns123abc: Big AI Safety Industrial Complex at it again\n\nQT @TaylorLorenz: Making anti AI content on YouTube ...
@altryne: RT @TaylorLorenz: Making anti AI content on YouTube is not only profitable bc you’ll get a zillion v...

18. simonguozirui (Group Score: 91.2 | Individual: 39.0)

Cluster: 4 tweets | Engagement: 906 (Avg: 69) | Type: Startup

RT @kickingkeys: You can just RL a coding model to paint with javascript btw https://t.co/4x5B81kjUh

See 3 related tweets

@dejavucoder: is this what they call "at the intersection of art and technology"?\n\nQT @kickingkeys: You can just...
@eisokant: I really love this work! Shows how far code can go\n\nQT @kickingkeys: You can just RL a coding mode...
@adonis_singh: i love this so much\n\nQT @kickingkeys: You can just RL a coding model to paint with javascript btw ...

19. mitsuhiko (Group Score: 88.2 | Individual: 35.5)

Cluster: 3 tweets | Engagement: 162 (Avg: 211) | Type: Tech

Another cool case of building hard things with AI now! https://t.co/4Aq0I9rrRi\n\nQT @tobi: Git at Scale (by cursor) has been one of the most interesting blog posts i've read in a while. It came right when I was frustrated with Shopify's internal git system.

As an exercise, I've implemented it over the weekend as open source. It's a single rust binary that you can point at any S3 type object store. It uses WAL and CAS primitives and requires no other data store.

It also implements bundle-uri so large git repos (like our mono) are very fast to download as a chain of static bundles. Also comes with basic familiar UX.

https://t.co/oEnWK3eIL4

See 2 related tweets

@andersonbcdefg: RT @tobi: Git at Scale (by cursor) has been one of the most interesting blog posts i've read in a wh...
@didier_lopes: The start of this podcast is @sama talking about how Tobi is always ahead of the curve, and then he ...

20. DataChaz (Group Score: 86.5 | Individual: 29.1)

Cluster: 4 tweets | Engagement: 53 (Avg: 71) | Type: Tech

Elon Musk just went on the record:

"Grok 4.6 is the #1 coding model right now and GrokBot will be the #1 agentic tool in 6-12 months."

To back it up, he confirmed over 70% of @SpaceXAI engineers are already using it to build self-learning agentic systems.

In this 45-minute session with those engineers, we actually get to see the playbook.

It is a crash course in modern AI architecture and worth more than a Stanford degree.

Watch it today.

Then read @0xCodez's brilliant 10-step teardown on how to actually build these systems yourself:

deploy agents on their own cloud computers teach workflows by simply demonstrating them and put specialist bots in a chat to coordinate

even @elonmusk RT'd it 👀↓\n\nQT @0xCodez: https://t.co/eDBs67VMq2

See 3 related tweets

@kentcdodds: RT @0xRafy: SpaceXAI engineer (ex-Cursor):

"Grok Bot was vibe coded in days. humans weren't reading...

@DataChaz: RT @DataChaz: Elon Musk just went on the record:

"Grok 4.6 is the #1 coding model right now and Gro...

@DataChaz: RT @DataChaz: spending my Monday deep in GrokBot.

deploying an entire army of AI agents as we speak...