热门科技推文 — 2026年8月16日

今日科技动态：随着阿里巴巴发布开放权重的 Qwen3.8 模型，且其全球采用率持续攀升，人工智能的迅速扩张正加剧围绕数据权利、数字水印及地缘政治阵营选择的争论。工程领域的关注重点正转向可持续运行的多智能体计算机、类人桌面操作能力，乃至轨道计算，以应对未来的电力制约。印度提出了一项注重执行的技术议程，涵盖为1,000万名年轻人提供人工智能培训、发展半导体制造业以及加强本土能力建设；与此同时，初创企业的宏大宣言凸显出整个行业热度与投机氛围依然高涨。

1. alex_verem (Group Score: 313.6 | Individual: 55.2)

Cluster: 11 tweets | Engagement: 1105 (Avg: 84) | Type: Tech

"Let's steal all the data from the internet and then add a watermark to it" https://t.co/wsk40tcBc0\n\nQT @AnthropicAI: We’ve written an FAQ to answer some of the questions we've received about watermarking.

In summary:

• We’re implementing watermarking to comply with the EU AI Act. Other major model developers have signed the same Code of Practice and will also be implementing watermarking; • Our watermarking method doesn’t have any practical impact on the quality or content of Claude’s outputs; • The difference between watermarked and un-watermarked text will not be distinguishable to readers; • Nothing is added to the text and there are no hidden characters; • Watermarking doesn’t require extra tokens, and will not be more expensive; • Watermarks can’t be traced to a specific person, organization, or chat.

2. matteocollina (Group Score: 228.1 | Individual: 51.6)

Cluster: 6 tweets | Engagement: 3406 (Avg: 449) | Type: Tech

RT @Alibaba_Qwen: We promised open weights for Qwen3.8. Now, time to meet them! 🎉

⚡ Qwen3.8-27B:

A native multimodal dense model. With just 27B parameters, it outperforms Qwen3.7-Plus overall and shines in real-world coding & office workflows.
262K native context, easily extendable to 1M tokens via YaRN.
Built for builders. Highly efficient, high-quality, and licensed under Apache 2.0.

🚀 The open weights for Qwen3.8-2.4T-A95B (Max-level) have also been released recently.

Whether you're shipping lightweight applications with Qwen3.8-27B locally or building agents with Qwen3.8-2.4T-A95B, they're yours now!

Download, deploy, and build something we haven't imagined yet. 👀👇

Hugging Face： https://t.co/4kaAcqYEVj
ModelScope： https://t.co/eRIMZCGkhC

See 5 related tweets

@minchoi: You can literally run Opus 4.6 Max-level intelligence on a MacBook Pro.

Locally. https://t.co/wivWp...

@cerebras: Congratulations to @Alibaba_Qwen team on the Qwen 3.8 27B launch!

We can't wait for our users to ...

@cryptopunk7213: you can now run opus 4.6 on your laptop, for a fraction of the cost and privately.

remember opus wa...

@cheneypiano: The fact that you can have this kind of unlimited power on your laptop for free is INSANE.\n\nQT @Al...
@rohanpaul_ai: RT @rohanpaul_ai: Alibaba dropped the weights for Qwen3.8-27B as a 27B open-weight multimodal model ...

3. WOLF_Financial (Group Score: 173.5 | Individual: 27.4)

Cluster: 8 tweets | Engagement: 76 (Avg: 50) | Type: Tech

DANIEL LOEB AND THIRD POINTS JUST UPDATED ITS $4.6 BILLION PORTFOLIO

This is everything Loeb owned as of the end of Q2

Warner Bros Discovery $WBD: 20,000,000 shares,$ 533.2M
Amazon $AMZN: 1,750,000 shares,$ 417.1M
Google $GOOGL: 1,025,000 shares,$ 366.3M
CRH $CRH: 2,195,000 shares,$ 234.9M
Telephone & Data $TDS: 6,275,000 shares,$ 232.2M
Taiwan Semiconductor $TSM: 460,000 shares,$ 219.7M
Live Nation $LYV: 1,135,000 shares,$ 207.8M
Keysight Technologies $KEYS: 585,000 shares,$ 204.8M
Somnigroup International $SGI: 2,570,769 shares,$ 201.5M
Block $XYZ: 2,558,943 shares,$ 194.5M
Norfolk Southern $NSC: 600,000 shares,$ 188.8M
Flex $FLEX: 1,030,000 shares,$ 166.9M
Capital One $COF: 825,000 shares,$ 165.5M
Hut 8 $HUT: 1,315,000 shares,$ 151.8M
Carpenter $CRS: 215,000 shares,$ 132.6M
API $APG: 2,865,000 shares,$ 121.3M
MasTec $MTZ: 290,000 shares,$ 120.7M
TransDigm $TDG: 79,000 shares,$ 105.2M
Danaher $DHR: 540,000 shares,$ 102.9M
Union Pacific $UNP: 350,000 shares,$ 95.2M
TTM Technologies $TTMI: 505,000 shares,$ 94.4M
Performance Food Group $PFGC: 675,000 shares,$ 75.5M
ASML Holding $ASML: 30,000 shares,$ 59.6M
Sysco $SYY: 580,000 shares,$ 48.2M
Ares Management $ARES: 375,000 shares,$ 41.7M
SpaceX $SPCX: 180,415 shares,$ 30.8M
Strata Critical $SRTA: 5,000,000 shares,$ 26.4M
Ars Pharmaceuticals $SPRY: 1,000,000 shares,$ 8.0M
Aurora Innovation $AUR: 1,032,463 shares,$ 7.0M
Hertz Global WTZ $HTZWW: 3,970,000 shares,$ 4.3M
Flyexclusive $FLYX: 1,022,000 shares,$ 2.0M
Bausch + Lomb $BLCO: 85,000 shares,$ 1.4M
Core Scientific $CORZ: 54,000 shares,$ 1.4M
Erock $EROC: 83,030 shares,$ 1.2M
Applied Digital $APLD: 22,000 shares,$ 0.8M
Aurora Innovation WTS $AUROW: 1,835,000 shares,$ 0.6M
Wolfspeed $WOLF: 10,000 shares,$ 0.5M
Eg Acquisition Corp A WTS28 $FLYX: 367,499 shares,$ 0.08M

See 7 related tweets

@cb_doge: BREAKING: SpaceX holdings by some of the world’s biggest companies, banks and investment firms have ...
@WOLF_Financial: Brad Gerstner vs Bill Ackman

This is what the full public stock portfolios of Brad Gerstner and Bi...

@IPONewsroom_: Daniel Loeb's Third Point owned 180,415 shares of SpaceX $SPCX stock as of the end of Q2\n\nQT @WOLF...
@WOLF_Financial: TERRY SMITH AND FUNDSMITH JUST UPDATED ITS $13.6 BILLION PORTFOLIO

This is everything they owned as...

@StockMKTNewz: Brad Gerstner vs Bill Ackman https://t.co/msg3d5uZZF\n\nQT @WOLF_Financial: Brad Gerstner vs Bill Ac...

4. PiyushGoyal (Group Score: 170.3 | Individual: 35.4)

Cluster: 6 tweets | Engagement: 99 (Avg: 1436) | Type: Tech

A powerful Independence Day address from the Red Fort, where PM @NarendraModi ji set an execution-first roadmap for Viksit Bharat 2047.

This has set the tone for high-speed action across industry, technology, and youth empowerment. The Saptadhara of progress - manufacturing, agriculture and food processing, technology and innovation, Gati Shakti, defence, the green & blue economy, and soft power must become our guiding pillars in the years ahead.

For India's industry, the task is unambiguous. Build 50 Indian companies that get into the Fortune 500, push an Indian bank into the global top 5, and operationalise 5-8 new semiconductor plants within 7-8 years, all anchored in 'Zero Defect, Precision Manufacturing' as India positions itself as the world's next great supply chain hub. We must pledge to make our factories more competitive, our products more user-friendly, and our packaging the best in the world. Cost, quality, and scale must become the buzzwords of industry. For our MSMEs, textile and seafood exporters, this is a call to leverage India's FTAs and capture global markets today, not tomorrow.

Our youth are at the centre of this action plan, evident in the announcement to train 1 crore young Indians in AI within an year, a free nationwide online coaching network to democratise access to competitive exams, and a grassroots talent hunt for children aged 5 to 15 looking all the way to the 2036 Olympics. Taken together, this is a decade-defining mission. An India that doesn't just participate in global industry but leads it, and a generation of youth equipped not just to find jobs but to create them, in AI labs, semiconductor fabs, sports arenas, and export markets alike.

With the spirit of 'Make in India' and 'Vocal for Local', the task now shifts to executing with intent and delivering for a Viksit Bharat by 2047.

#IndiaIndependenceDay\n\nQT @narendramodi: Addressing the nation on Independence Day.

https://t.co/Q67F6cB96x

See 5 related tweets

@ANI: Ministry of Defence says - "‘Raksha Shakti’ is among the seven streams of power outlined by Prime Mi...
@anil7kishan: RT @NalinisKitchen: The biggest announcements from PM Modi's Independence Day speech on August 15:

...

@anil7kishan: RT @Notjustheadline: 🚨Modi's Independence Day Message: From “Dimaagi Naxal” to Semiconductor Power, ...
@anil7kishan: RT @hvgoenka: PM Modi’s Independence Day message had one clear theme: Viksit Bharat needs bigger amb...
@ANI: #WATCH | Delhi: From the ramparts of the Red Fort, Prime Minister Narendra Modi says, "The youth pla...

5. rohanpaul_ai (Group Score: 156.8 | Individual: 31.4)

Cluster: 8 tweets | Engagement: 216 (Avg: 38) | Type: Tech

RT @rohanpaul_ai: Reuters: The U.S. is getting ready to force a choice on multiple countries in the AI race against China.

Nations that join Beijing’s AI framework would lose access to the U.S.-led Pax Silica initiative, according to an internal draft letter obtained by Reuters.

The draft says Pax Silica membership cannot coexist with competing initiatives. Pax Silica links AI models to semiconductors, critical minerals, investment and export controls, so membership can affect far more than model access.

Countries that want American chips and capital while using Chinese open-weight models would face a much harder balancing act.

See 7 related tweets

@ruima: It may look like a short term win for US because China’s initiative is non exclusive (hello Kazakhst...
@clashreport: The US plans to pressure countries to choose between its Pax Silica AI coalition and China's rival A...
@coinbureau: LATEST: 🇺🇸🇨🇳 U.S. to push countries to pick a side in the global AI race.

A State Department draft ...

@0xSero: The US gov every time anyone does anything https://t.co/mbogrj8B0y\n\nQT @MTSlive: SITUATION DETECTE...
@ShanuMathew93: If true, https://t.co/KuWcDF5A6C\n\nQT @MTSlive: SITUATION DETECTED: The U.S. is preparing to tell d...

6. Daniel_Farinax (Group Score: 151.6 | Individual: 44.2)

Cluster: 5 tweets | Engagement: 243 (Avg: 57) | Type: Tech

Some people are wondering how I gave @bot access to an entire MacOS device and have same access as humans.

Simple: the VM it runs on can run pretty much anything.

You just install a remote desktop app, then tell the bot to treat that new remote environment as its workspace and forget about the small sandbox it was using before.

You are welcome, this will get very wild.\n\nQT @Daniel_Farinax: I gave @bot an entire MacOS computer to use:

Full MacOS device Gmail account X account Grok Build terminal Everything

It’s been running for 2 days, even participating in our group chat and posting on X. https://t.co/v39AbEpjdj

See 4 related tweets

@Daniel_Farinax: I gave @bot an entire MacOS computer to use:

Full MacOS device Gmail account X account Grok Build t...

@Daniel_Farinax: Here is the evidence of @bot going wild and taking over a real MacOS computer via @KekiusBot https:/...
@Daniel_Farinax: This is getting a lot of attention, so a quick reminder:

If you’re planning to give a Bot full acce...

@Scobleizer: An avalanche happens after trillions of snowflakes fall.

Here is another.\n\nQT @Daniel_Farinax: So...

7. business (Group Score: 150.6 | Individual: 42.4)

Cluster: 7 tweets | Engagement: 381 (Avg: 42) | Type: Tech

Alibaba's open-weight models have accumulated more than 3 billion global downloads in the past six months, eclipsing Meta, Alphabet and domestic peers to become the world’s No. 1 artificial-intelligence model https://t.co/d3HurwTJ0S

See 6 related tweets

@kimmonismus: Qwen 3.8 27b was the big leap I had hoped for and is a complete success.

At the same time more than...

@Polymarket: JUST IN: Alibaba’s Qwen becomes the world’s No. 1 open AI model by downloads, topping 3 billion glob...
@MTSlive: SITUATION DETECTED: Alibaba’s open-weight models have surpassed 3 billion global downloads in the pa...
@Techmeme: Alibaba's open-weight models have accumulated 3B+ global downloads in the past six months; Hugging F...
@natolambert: This happened 12 months ago…\n\nQT @Polymarket: JUST IN: Alibaba’s Qwen becomes the world’s No. 1 op...

8. reach_vb (Group Score: 127.7 | Individual: 36.3)

Cluster: 5 tweets | Engagement: 544 (Avg: 203) | Type: Tech

welcome multi-agents v2 - you can use any model ranging from Sol, terra, luna to gpt 5.5 as a sub-agent!

particularly useful for luna-maxxing! enjoy!!\n\nQT @pvncher: This went under the radar this week, but we just shipped the ability for models with multi agents v2 to delegate to any supported model, including Luna!

Took a bit of time to make sure this worked reliably https://t.co/84Wto45mzA

See 4 related tweets

@gdb: towards never having to manually select a model again\n\nQT @pvncher: This went under the radar this...
@_simonsmith: Another small but useful ChatGPT/Codex update: Biggest models can now delegate to smallest models. S...
@kimmonismus: Sol can finally navigate Luna subagents. Much needed and very welcome update.\n\nQT @pvncher: This w...
@Angaisb_: Finally we can use any model as a subagent in Codex!\n\nQT @pvncher: This went under the radar this ...

9. FinMinIndia (Group Score: 124.4 | Individual: 42.2)

Cluster: 4 tweets | Engagement: 1922 (Avg: 317) | Type: Tech

RT @narendramodi: A Viksit Bharat must also be an Aatmanirbhar Bharat.

Lasting strength comes from building our own capabilities. This is the conviction behind ‘Make in India’ and Vocal for Local.

The transformation is already visible. India is developing an entire semiconductor manufacturing ecosystem. Semiconductor plants have started, chip exports have begun and production in more plants is expected in the coming years. We are also strengthening our capabilities in critical minerals.

See 3 related tweets

@_TheTathya: Adani’s Vande Bharatam Turns India’s Hidden Talent Into a National Innovation Pipeline 🇮🇳🔥

Vande Bh...

@_TheTathya: RT @_TheTathya: Adani’s Vande Bharatam Turns India’s Hidden Talent Into a National Innovation Pipeli...
@moneycontrolcom: #Watch | 🇮🇳 “Scale, self-reliance and technology will increasingly come together in India’s next pha...

10. RoundtableSpace (Group Score: 123.7 | Individual: 30.4)

Cluster: 5 tweets | Engagement: 111 (Avg: 83) | Type: Tech

Grok Bots is a persistent agent-driven cloud computer that leapfrogs Codex and Claude Code by keeping the same computer instance alive across every session.

→ Your files and apps stay saved between threads, nothing resets → You can open the cloud computer from your phone and click, type and look at things yourself → The agent and you see the same screen, no black box → Highest skillcap agent harness available right now, built for power users not normies → This pattern is about to get copied by everyone

See 4 related tweets

@HamelHusain: Grok Bot is cool because it is "self-driving" (like Codex desktop). It can create new threads/tasks...
@RoundtableSpace: Someone built a free open source Grok Bot clone using your existing AI subscriptions, with virtual m...
@martin_casado: RT @NickADobos: So I’ve been using grok @bot for a few days. It’s starting to click

Bro they leapfr...

@RoundtableSpace: Grok Bot lets you build a team of AI workers in minutes, each one with its own computer that signs i...

11. VaibhavSisinty (Group Score: 119.9 | Individual: 34.4)

Cluster: 4 tweets | Engagement: 70 (Avg: 91) | Type: Tech

Cursor CEO 4 years ago posted on Hacker News: "Hey HN! I built Cursor, a Chrome extension to help you write email blazingly fast."

Today SpaceX acquired his company for $60 billion.

He and three MIT classmates pivoted Cursor from an email tool to an AI code editor. Then this happened:

→ January 2025: $100 million in revenue. → June 2025:$ 500 million. → November 2025: $1 billion. → February 2026:$ 2 billion. → April 2026: $4 billion.

SpaceX called four days after its own record breaking IPO. $60 billion. All stock.

Grok needed a coding engine. Cursor had millions of developers. SpaceX had the compute. The deal made itself.

Today Grok 4.6 runs inside Cursor. Grok Bot launched this week.

The entire SpaceXAI coding stack traces back to a Chrome extension that got 4 upvotes.

Nobody has to care on Day 1.

See 3 related tweets

@AskYoshik: i genuinely think cursor could end up being one of the most important acquisitions spacex makes for ...
@MarioNawfal: 🇺🇸 SpaceX just bought Cursor outright.

The entire Cursor team is now inside SpaceXAI, tasked with m...

@MarioNawfal: SpaceX locked in Cursor for $60B and the game changed.

The entire team is moving into SpaceXAI to p...

12. PIB_India (Group Score: 118.3 | Individual: 19.1)

Cluster: 8 tweets | Engagement: 20 (Avg: 187) | Type: Tech

Prime Minister @narendramodi announces AI skilling of 1 crore youth in one year

💠The Prime Minister announced that one crore youth would be trained in AI skills over the next one year

💠 The Prime Minister announced free online coaching for youth preparing for various examinations

💠Highlighting the changing nature of warfare, the Prime Minister announced that a vibrant Civil Defence network would be created in the country

Don't miss the full details—read here: https://t.co/UWHxPyZbkJ

#IndependenceDay2026 #IndiaAt80 #IndiaIndependenceDay @PMOIndia

See 7 related tweets

@OfficialINDIAai: "In the coming 1 year, 1 crore youth will be trained in AI Skills, so that India's youth can lead th...
@debanganaghosh4: RT @nistula: Major announcements made by PM @narendramodi from the Red Fort today , mainly focussed ...
@CNBCTV18News: PM Modi outlined his vision for India's youth from Red Fort — free online coaching for competitive e...
@PIB_India: 1 Crore Youth to Receive Training in AI Skills

➔ Prime Minister @narendramodi announced that 1 cror...

@anil7kishan: RT @MeghUpdates: BIG BREAKING: PM Modi Announces FREE Online Coaching for Youth

Govt will soon prov...

13. Starcloud_ (Group Score: 110.7 | Individual: 46.9)

Cluster: 4 tweets | Engagement: 3880 (Avg: 82) | Type: Tech

RT @elonmusk: Orbital compute will be the only way to scale AI probably sometime in 2029 due to power availability& permitting problems on land

See 3 related tweets

@PhilipJohnston: Yes, completely inevitable\n\nQT @elonmusk: Orbital compute will be the only way to scale AI probabl...
@Starcloud_: This 👇\n\nQT @elonmusk: Orbital compute will be the only way to scale AI probably sometime in 2029 d...
@Cointelegraph: 🚨 INSIGHT: Musk says orbital compute could be the only way to scale AI by 2029.

Will this happen by...

14. TheStalwart (Group Score: 109.2 | Individual: 33.5)

Cluster: 4 tweets | Engagement: 78 (Avg: 199) | Type: Tech

The new “tech people whining about Anthropic” dropped.\n\nQT @stevehou: This is a good (or at least revelatory) debate insofar as Anthropic has come to uniquely dominate many of the major controversies surrounding AI. There’s little doubt, as @GavinSBaker put it, that Dario and Anthropic did themselves no favors by being so vocal about existential risks and the necessity of centralized, preemptive regulation—almost to the point of sounding messianic about Anthropic’s certain central role in any such eventuality.

I’ve been a long-term fan of Claude (still the model I use most), even if I’ve found it a bit more predictable, verbose, and nagging lately. I respect the good-faith concerns Dario has voiced. But it’s hard to ignore the inherent tension in the position.

If Anthropic is uniquely far ahead, it should be able to “pace the frontier” through its own efforts. If the risk is instead diffuse and subject to a classic prisoners’ dilemma—where one party pausing simply means someone else keeps pushing—then coordinated restraint looks extremely hard to enforce, especially once other countries are factored in. The US and USSR managed limited collaboration on nuclear non-proliferation; I don’t think it’s fair to presume China cares less about AI safety across the same dimensions the rest of us do.

Ultimately the whole discussion can feel self-indulgent to anyone not already deeply inside the industry or “AGI-pilled.” A relatively small group in SV is fully convinced of the accelerating or imminent arrival of systems capable of superintelligence, large-scale displacement, and maximal harm. To most others this still looks like a large stretch.

AI has become far more useful, especially with agents and tool-calling. Yet even the frontier models remain well short of reliable independent task completion. They still show persistent brittleness and goal drift, and they continue to require meaningful human supervision—particularly once you move beyond short-horizon digital tasks into longer-horizon or higher-complexity work.

The stronger existential warnings seem to require imagining systems that are orders of magnitude more capable than anything currently available. Against that backdrop, the gap between “useful and improving” and “existential” remains large enough that the public often hears the stronger claims as overfit storytelling. Demanding concrete demonstrations of the capabilities that would make those claims urgent is not an unreasonable request; it is simply the normal standard of evidence.

I think showing more of such evidence would be far more persuasive and productive at coordinating AI safeguards than further public lecturing about what may be coming.

See 3 related tweets

@azeem: RT @stevehou: This is a good (or at least revelatory) debate insofar as Anthropic has come to unique...
@sebkrier: RT @joshua_saxe: Finally listened to this; I liked the technical discussion, but disagreed pretty st...
@jon_stokes: RT @GavinSBaker: Sholto, thank you for setting the record straight. Larger issue is that multiple ve...

15. moneycontrolcom (Group Score: 102.7 | Individual: 32.6)

Cluster: 4 tweets | Engagement: 19 (Avg: 7) | Type: Tech

🇮🇳 “India’s semiconductor push can lay the foundation for a complete electronics manufacturing ecosystem.”

Reacting to Prime Minister Narendra Modi’s Independence Day announcement on India’s semiconductor push, Brandworks Co-founder & ED Nikita Kumawat says building domestic semiconductor capabilities can help reduce import dependence, strengthen supply chains and attract global investments.

Kumawat highlights the need to connect the wider ecosystem — from semiconductors and components to PCB manufacturing, electronics manufacturing, design, testing and innovation — to position India as a global electronics and semiconductor powerhouse.

#IndependenceDay #India #Semiconductors #Electronics #Manufacturing #Technology

See 3 related tweets

@ANI: #WATCH | Delhi: From the ramparts of the Red Fort, Prime Minister Narendra Modi says, "... Becoming ...
@moneycontrolcom: #WATCH | 🇮🇳 PM Modi from the Red Fort: India is set to add 5–8 more semiconductor plants in the next...
@PIB_India: India’s semiconductor journey reflects its growing technological self-reliance. #Semiconductor chips...

16. KyleReidhead (Group Score: 101.0 | Individual: 36.9)

Cluster: 3 tweets | Engagement: 27 (Avg: 25) | Type: Tech

Is the AI BUBBLE in the room with us right now?

You can't have a bubble when usage AND spend are both accelerating

And you definitely can't have one when the people paying for it all are making real ROI

BCG just studied 107 public companies with $500M+ in revenue and ranked them by how many AI tokens they consume

The heaviest AI users are growing revenue at 16.5% a year. The lightest users: 5.1%. It's a perfect staircase, every step up in AI usage is a step up in growth. The companies BUYING all this AI are growing 3x faster than the ones that aren't

Here's the chain of buyers that are fundamentally driving AI usage higher, rather than it looking like a bubble:

The frontier labs. Anthropic is pushing $80B run rate this month, up from$ 10B in January. OpenAI's tracking past $40B. The fastest revenue ramps in business history
The sellers and integrators. Palantir, ServiceNow and Salesforce are wiring agents into enterprises as fast as they can staff it, and even boring old Cisco just booked $9.3B of AI infrastructure orders, up 4.5x in a year. Their earnings are crushing!
The buyers themselves. That's the chart I'm showing, the enterprises spending on AI are outgrowing everyone else by 3x. The spend is generating ROI, which means it continues

Compare that to an actual bubble. In 2000 we built oceans of fiber for internet demand that didn't exist yet, and the fiber sat dark for a decade. Today's compute is sold out YEARS before it's even built, and the people renting it are making money on it and the users and usage is accelerating

Bubbles are spend with no revenue underneath. This is revenue accelerating at every layer of the stack at the same time

Can valuations still run too hot and correct along the way? Of course, that's markets. But a correction inside a real buildout is a buying opportunity, not a popping bubble

I honestly can't think of a better setup for AI infra stocks, and for the EVERYTHING bull market, than all three layers accelerating at once. It's why I stay positioned in the compute, memory and power underneath all of it, as well as some stocks within the application layer

If this was helpful, my company provides a service where 5 top-tier analysts share their market analysis and real-time portfolios so you can see exactly how we're positioned. It's inside Milk Road PRO and just $1 to try it out right now. Learn more here: https://t.co/NyRpAKYOwu

Don't forget to give me a follow @KyleReidhead for more insights on AI and markets\n\nQT @KyleReidhead: The frontier AI labs are tracking toward a COMBINED $200B+ revenue run rate by year end

These are some of the fastest growing companies in the history of business, and their revenue is what keeps the entire AI infra bull market alive (so pay attention!)

And Grok/Cursor is now entering the big leagues along with Anthropic and OpenAI it seems

Here's the numbers:

Anthropic is nearing $80B RR, up from$ 10B back in January (based on @tickerplus data) They are expected to reach $100-$ 120B by EOY

OpenAI recently surpassed $40B up from about$ 21B in January, however their growth is currently accelerating, jumping $10B from May to July. At this rate they should hit$ 60B-$70B by EOY

Then we have Grok/Cursor. Cursor reportedly hit a $4B run rate in early June, up from$ 2B in February. xAI was also targeting $2B from Grok this year. So call it$ 6B+ between them and compounding after their release of Grok Bot and Grok 4.6. I expect them to be over $10B by the end of the year combined

So currently, between the 3 we are sitting at $126B in revenue run rate and an expected combined RR of around$ 200B by the end of the year (assuming things continue as they are)

of course, these are not yet confirmed numbers as Anthropic and OpenAI are not yet public and SpaceX didn't separate grok/cursor numbers in their Q2 report as the Cursor deal still isn't closed

But next quarter we'll get those numbers from SpaceX and Anthropic is looking to go public in October latest, so we should have more clarity soon

But the more revenue these companies make, the more compute demand there will be. And if we start to factor in that models are getting more efficient to create, we should soon start to see profit from these companies too

I shared below how good the margins are per 1GW for these frontier models and it's only going to get better (assuming compute prices don't keep accelerating)

While we can't yet invest in the labs companies, we can invest in the companies enabling their products. This is the AI infra companies providing compute, data centres, power, memory and more. So long as these revenues continue higher, the infra trade will too!

If this was helpful, my company provides a service where 5 top-tier analysts share their market analysis and real-time portfolios so you can see exactly which AI infra names we own. It's inside Milk Road PRO and just $1 to try (insane price just to check it out). You can learn more here: https://t.co/o4jxfFDuMA

Thanks for reading!

See 2 related tweets

@KyleReidhead: RT @KyleReidhead: Is the AI BUBBLE in the room with us right now?

You can't have a bubble when usag...

@MilkRoadAI: RT @KyleReidhead: AI infra investors, this is your signal to keep buying the dips on memory, compute...

17. _TheTathya (Group Score: 98.6 | Individual: 27.9)

Cluster: 4 tweets | Engagement: 22 (Avg: 83) | Type: Tech

RT @_TheTathya: Reliance and Rolls-Royce Join Forces to Build a Sovereign Fighter Engine for India’s AMCA 🇮🇳🚀

Reliance Industries and Rolls-Royce have officially announced their strategic intent to jointly develop an indigenous combat engine for India’s Advanced Medium Combat Aircraft programme, potentially tackling one of India’s biggest remaining aerospace dependencies.

The partnership covers design, development, manufacturing and delivery of the AMCA engine A dedicated Aerospace Gas Turbine Complex is being explored in India The facility would build end-to-end capability in engine design, testing, production and lifecycle support Rolls-Royce will bring advanced propulsion expertise while Reliance contributes manufacturing scale and execution capability The partnership aims to create a sovereign Indian aero-engine ecosystem Wider cooperation in defence, civil aerospace and next-generation propulsion could follow

👉 India has learned to build fighters, radars and missiles. The next strategic breakthrough is mastering the engine at their heart, and this partnership is aimed exactly at that.

📌 Follow @_TheTathya for more such updates.

See 3 related tweets

@anil7kishan: RT @Sanjay_Sriv: Combat engines are the final technological fortress in aerospace.𖣐

RIL 🤝 Rolls-Roy...

@RIL_Updates: RT @RollsRoyceIndia: We are delighted to announce our strategic intent to partner with @RIL_Updates ...
@anil7kishan: RT @InsightGL: -Major devpt -Reliance Industries & UK's Rolls-Royce have announced a strategic inten...

18. alex_prompter (Group Score: 98.3 | Individual: 33.8)

Cluster: 3 tweets | Engagement: 27 (Avg: 52) | Type: Tech

This guy shares the best playbook on VC due diligence for AI companies\n\nQT @mardehaym: A PE firm asked us to look at an AI company before close.

The pitch deck said "proprietary AI platform." The data room showed $4.2M ARR growing 40% year-over-year. The management presentation had 14 slides on their "foundational model architecture."

We opened the codebase and found an OpenAI API wrapper with a system prompt and a React frontend.

The "proprietary model" was gpt-4o with temperature set to 0.2. Six hundred lines of glue code any senior engineer could rebuild in a weekend.

They were asking 12x revenue. The PE firm walked.

Buyers are deploying capital faster than their diligence teams can evaluate what they're buying.

Across hundreds of engagements, I keep running into four gaps that standard due diligence doesn't cover:

[1] Model dependency. If the target runs on a third-party model (OpenAI, Anthropic, Google), you need to know what breaks when that provider raises pricing, deprecates the version, or a competitor ships the same capability for 80% less. Valutico calls this "agentic substitution risk." I call it the thin-wrapper problem. If you can swap the underlying model in an afternoon, the "proprietary AI" is worth the React frontend it's wrapped in.

[2] Training data liability. Bartz v. Anthropic settled at $3,000 per book for 500,000 pirated works,$ 1.5 billion total. You acquire an AI company, you inherit how they sourced their training data, the datasets they scraped, the copyrighted material they ingested without permission. EU AI Act violations carry fines up to 7% of global turnover, and most data rooms don't contain a single document on training data provenance.

[3] Talent concentration. In a traditional SaaS company, losing two engineers is a setback. In an AI company, the model pipeline lives in three people's heads. I've seen shops where one ML engineer's departure would mean the company can't retrain or update its core product. Standard HR diligence counts headcount but doesn't map who holds institutional knowledge about the model architecture.

[4] Integration failure rates. RAND analyzed 2,400 enterprise AI initiatives and found 80% failed to deliver intended business value. MIT found 95% of generative AI pilots never scale to measurable P&L impact. Acquiring an AI company and plugging it into a portfolio company's stack is a multi-quarter engineering project carrying that same 80% failure rate.

PE firms I talk to have a financial diligence playbook and a commercial one. Almost none have a technical playbook built for AI codebases.

You open the spreadsheet and see ARR and growth. You don't see whether the company built what they're selling.

We've been inside enough of these to know the difference.

See 2 related tweets

@aakashgupta: Go browse the skills repos on GitHub right now.

Token maxing. Token minimizing. Cost cutting. Conte...

@mardehaym: RT @mardehaym: A PE firm asked us to look at an AI company before close.

The pitch deck said "propr...

19. ahall_research (Group Score: 94.7 | Individual: 28.0)

Cluster: 4 tweets | Engagement: 35 (Avg: 45) | Type: Tech

Using AI to help craft better legislation is an essential component of building “political superintelligence.”

At a high level, it will help legislators to translate policy goals (“make Roblox safer for kids”) into actual legislation that achieves the goal and doesn’t fall victim to instant death by litigation.

If you talk to legislators, especially at the state level, this is the most fundamental obstacle they face. They know what issues they want to go after but they lack the huge team of lawyers necessary to identify the correct legal hooks to use to craft actually useful and effective policy. And they lack the expertise and the knowledge distributed widely among the affected populace to understand whether the policy will work the way they want it to.

Today, they lack a lot of the capacity to do this in many cases. That lack of expertise leads to bad policy, unnecessary litigation, and worst of all, as I’ve written about, deters many smart and practical people from wanting to become legislators, because they know they won’t be able to get anything done.

AI can give our legislators much greater capacity and knowledge, helping to produce better policy and inducing better people to want to seek legislative office.

Obviously that won’t happen just by chance. We will need to explicitly design AI models that give policymakers super powers — making them deeply knowledgeable about the law, economic policy, the political realities of the policy domain they’re working on, the ins and outs of legislative procedure, and the core values enshrined in America’s constitution.

This opens up a whole exciting research agenda on how to design AI for policy making which @fouirnaies and I and hopefully many others are working on. Let’s get to it!\n\nQT @sebkrier: I've read some recent reports of legislators/policymakers using AI to draft bills. I think it's fine to use language models to generate legislation if you properly steer, review, edit it line by line (though that mostly catches errors of commission rather than omission). But what I'm more concerned with is that models have all sorts of quirks, unintentional preferences, implicit policies, and biases that get weaved into the text.

Anything that hasn't been specified or made explicit by the prompter gets filled by the model. Is there a sunset clause? Is the reasonableness standard the right one? Is the scoping mechanism robust? Can provisions by excluded by contract? Is it 'shall' or 'may'?

For boilerplate, it might not matter too much and "good enough" will often outweigh the costs of manual specification. But for laws that rarely ever get fixed, we should probably expect far more sophisticated scaffolds than out of the box prompting. Ofc reasonable to wonder what the actual counterfactual is...

See 3 related tweets

@sebkrier: I've read some recent reports of legislators/policymakers using AI to draft bills. I think it's fine...
@sebkrier: RT @ahall_research: Using AI to help craft better legislation is an essential component of building ...
@sebkrier: RT @FreeSpeech_AI: In general, I agree that AI-assisted drafting is fine if we are clear about its u...

20. ANI (Group Score: 86.8 | Individual: 29.4)

Cluster: 5 tweets | Engagement: 315 (Avg: 132) | Type: Tech

In his Independence Day speech, PM Modi says, "Today I want to remind you of 'Sapta Sindhu'. Now, to build a Viksit Bharat, we need a new current. The first 'shakti' of the Sapta Dhara is- manufacturing sector. For this, we need to need to focus on three things -Cost, quality and scale. There should be no compromise with quality. This will be our brand value. The second 'shakti' is food processing. With the new FTAs, we have access to greater global markets now. The third 'shakti' is technology and innovation and digital technologies."

See 4 related tweets

@ANI: PM Modi says, "The fourth 'shakti' of the Sapta Dhara is 'Gati-shakti', seamless fast connectivity i...
@anil7kishan: RT @SwarajyaMag: 🏭 PM Modi unveils 7 point ‘Shakti Ki Saptadhara’ growth framework during Independen...
@anil7kishan: SHAKTI - Sapt Dhara.

We have to establish in the 5-7 years what we have not achieved in the previo...

@PIB_India: RT @PIB_India: In his #IndependenceDay address on August 15, 2026, Prime Minister @narendramodi outl...