科技热门推文——2026年8月24日

今日科技要闻：人工智能的普及进程正与经济和社会现实发生碰撞，与此同时，英伟达、阿里巴巴和 xAI 正加大对开放模型、基础设施及智能体编程领域的投资。据报道，Hugging Face 正吸引潜在收购方的兴趣，其估值超过 130 亿美元；初创企业则在探索共享式人工智能架构和更严谨的评估方法。此外，针对数据中心的反对声浪日益高涨；有报道称，伊朗发动的攻击导致英国一家发电厂运营受阻，进一步加剧了网络安全担忧。印度也在积极吸引全球太空领域的创业者。

1. zerohedge (Group Score: 243.1 | Individual: 45.0)

Cluster: 8 tweets | Engagement: 2585 (Avg: 722) | Type: Tech

Turns out you can't change human behavior and decades of habit to suit your IPO timeline\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 7 related tweets

@zephyr_z9: "I think I was wrong about a few things, but one in terms of the speed: the economy just has so much...
@firesidealpha: Sam Altman says OpenAI shouldn't compete with all its customers or subsume the economy

"But that's ...

@KatieMiller: Sam Altman blows whichever way the wind does\n\nQT @firesidealpha: Sam Altman admits he was wrong on...
@edzitron: Oh ok lol\n\nQT @firesidealpha: Sam Altman admits he was wrong on AI's timeline and says society and...
@TaylorLorenz: “We've all been too ambitious on timelines. Even with this incredible technology, society and the ec...

2. ANI (Group Score: 206.1 | Individual: 62.7)

Cluster: 7 tweets | Engagement: 3609 (Avg: 136) | Type: Tech

#WATCH | Delhi: During interaction with CEOs and Founders of Space Startups, Prime Minister Narendra Modi said, "...I believe we should create an aura where people from all over the world feel that if they want to build a life in this field, they should come to India—join an Indian startup or the Indian space agency. I am convinced that we can attract this entire talent pool like a magnet, and it can become a tremendous strength for us...We are highly cost-competitive. There is no question regarding the quality of talent, and since the people of our country naturally possess a high capacity for risk-taking, we can rapidly carve out a niche for ourselves."

(Source: DD)\n\nQT @ANI: #WATCH | Delhi: During interaction with CEOs and Founders of Space Startups, Prime Minister Narendra Modi said, "You have shown courage in a new sector. The most important thing is that you placed your trust in the government. My view has always been that whatever decisions we make, we must remain consistent. We shouldn't keep changing them constantly; that way, anyone willing to take a risk can feel assured—knowing that if something goes wrong due to their own error, that’s one thing, but no one will abandon them. That is the trust we are striving to build."

See 6 related tweets

@mygovindia: Think India’s space story is just about rockets? Think again.

Under PM @narendramodi’s leadership,...

@anil7kishan: RT @pradip103: PM @narendramodi Ji :

‘India Can Become a Global Space Talent Magnet 🔥

Addressing ...

@moneycontrolcom: #India | PM Modi wants India to create an ‘aura’ that attracts global space talent
PM Modi: India...
@mygovindia: India Has What Space Needs!

While interacting with Space Startup CEOs, PM @narendramodi shared that...

@_TheTathya: RT @_TheTathya: PM Modi Backs India’s Private Space Start-Ups, Pushes for Global Commercial Leadersh...

3. MTSlive (Group Score: 192.8 | Individual: 49.8)

Cluster: 6 tweets | Engagement: 1431 (Avg: 62) | Type: Tech

SITUATION DETECTED: Nvidia is using its $6 billion deal with Poolside to develop one of the world’s most powerful open-weight models, aiming to challenge both Chinese labs and the US frontier. More than 100 Poolside engineers will be joining the Nemotron project, per WSJ.

See 5 related tweets

@PolymarketMoney: BREAKING: Nvidia is making one of its biggest open-weight AI moves yet, paying $6,000,000,000.00 to ...
@ollama: .@poolsideai engineers were awesome to collaborate with to launch open models.

Excited for what’s ...

@WSJ: Nvidia is planning to use a $6 billion deal it struck this week to build one of the world’s most pow...
@ctnzr: RT @ollama: .@poolsideai engineers were awesome to collaborate with to launch open models.

Excited...

@Techmeme: Sources: Nvidia plans to use its $6B deal with Poolside to build an open-weight AI model to compete ...

4. 0xCodez (Group Score: 149.2 | Individual: 48.0)

Cluster: 5 tweets | Engagement: 2093 (Avg: 392) | Type: Tech

Elon Musk:

"in 6-12 months GrokBot will be #1 tool for building agentic systems. In a year 100% of code will be done by LLM's

at SpaceXAI, more than 70% of engineers already using GrokBot to build self-learning agentic systems"

In a 40-minute talk, Elon with SpaceXAI engineers discuss how the future of AI engineering will look like

worth more than 2 hours of Stanford lecture on AI engineering

watch today, then read the article below on building a self-learning agentic system with GrokBot\n\nQT @0xCodez: https://t.co/eDBs67VMq2

See 4 related tweets

@0xRafy: Elon Musk:

"in 6-12 months GrokBot will be #1 tool for building agentic systems. In a year 100% of ...

@0xCodez: Elon Musk:

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

@0xRafy: RT @0xRafy: Elon Musk:

"in 6-12 months GrokBot will be #1 tool for building agentic systems. In a y...

@0xCodez: RT @0xCodez: Elon Musk:

"Grok 4.6 is the #1 coding model right now and GrokBot will be #1 agentic t...

5. MollySOShea (Group Score: 142.6 | Individual: 42.7)

Cluster: 4 tweets | Engagement: 268 (Avg: 61) | Type: Tech

RT @davidsenra: My conversation with Sam Altman (@sama), co-founder & CEO of @OpenAI.

0:00 Tobi Lütke, AI-Native Companies & Why Adoption Moves Slowly 5:45 Sam's Own Resistance to AI & the Missing iPhone Moment 10:00 Models, Compute, Power Laws & Non-Consensus Talent 18:37 From AI-Obsessed Kid to Founder, Investor & Back Again 23:19 Impossible Problems, Scientific Discovery & Human Connection 30:16 AI's Two Biggest Risks: Loss of Control & Centralized Power 33:09 Iterative Deployment, AI Safety & Learning From Reality 40:27 Why People Fear AI & the Coming Small-Business Boom 46:13 Context, Memory & the Next Way We Will Work With AI 49:17 OpenAI's Platform Strategy & Killing Good Ideas 53:20 Peter Thiel, Paul Graham & the Value of Nonlinear Thinkers 1:00:42 How Y Combinator Changed Startups & Shaped OpenAI 1:04:03 Learning More From Success & the Power of Repetition 1:09:45 Building OpenAI Without Customers, a Product or a Playbook 1:15:57 Letters to His Son & Preserving the Story

Includes paid partnerships.

See 3 related tweets

@AndrewCurran_: Sam Altman on how he feels the industry has not done a very good job with messaging. https://t.co/oR...
@Hesamation: Sam Altman throws some pretty obvious punches at Dario Amodei when he admits the AI industry has a B...
@rohanpaul_ai: OpenAI does not want to become the application layer of AI; it wants to be the platform that other a...

6. rohanpaul_ai (Group Score: 122.6 | Individual: 34.0)

Cluster: 7 tweets | Engagement: 31 (Avg: 40) | Type: Tech

Alibaba is seeking $10.2B from a share sale after quarterly capex reached almost$ 10B amid its AI buildout.

Says 100% of net proceeds from the proposed HK$80B placement will fund AI capabilities, including infrastructure expansion.

Will make it the largest primary follow-on offering ever by a Hong Kong-listed company if completed.

Alibaba spent $11.9B buying back stock in fiscal 2025 and cut shares outstanding by 5.1%; now it wants$ 10.2B by issuing new shares for AI. Those decisions don't cancel each other, but they show how quickly capital priorities have changed.

Because infrastructure consumes cash before the resulting capacity is fully monetized, Alibaba is accepting dilution rather than funding the entire next leg from its balance sheet.

Because Alibaba owns Qwen, cloud infrastructure, and proprietary AI chips, it has several ways to earn from AI demand even if model pricing keeps falling.

See 6 related tweets

@coinbureau: BREAKING: Alibaba is raising $10.2 BILLION in a share placement and putting 100% of it into AI, per ...
@Reuters: Alibaba plans $10 billion Hong Kong share placement to fund AI spending https://t.co/ZhGRXw04gO http...
@Cointelegraph: 🇭🇰 UPDATE: Alibaba plans to raise $10.2B through a Hong Kong share placement to fund its global AI p...
@Techmeme: Alibaba plans to raise ~$10B in a follow-on share offering to fund AI investments; sources: it plans...
@FirstSquawk: ALIBABA SEEKS $10.2 BILLION IN HONG KONG SHARE SALE TO FUND AI EXPANSION, ITS LARGEST-EVER FOLLOW-ON...

7. StockMKTNewz (Group Score: 102.9 | Individual: 23.1)

Cluster: 5 tweets | Engagement: 61 (Avg: 225) | Type: Tech

Hugging Face is reportedly looking to get acquired at a valuation of at least $13 billion\n\nQT @IPONewsroom_: Hugging Face has been fielding acquisition interest at a valuation of at least$ 13 billion, according to Business Insider.

The AI developer platform has reportedly been working with a bank to gauge potential bidders, though no deal has been reached.

Hugging Face was last valued at $4.5 billion in 2023, meaning a$ 13 billion deal would value the company at nearly 3x its previous valuation.

Unlike frontier model developers such as OpenAI or Anthropic, Hugging Face sits at the infrastructure layer of the AI ecosystem, allowing developers to publish, share, download and build with models from across the industry.

The reported interest comes shortly after Stripe agreed to acquire AI model marketplace OpenRouter for around $8 billion, highlighting how aggressively buyers are valuing platforms that sit between AI models and developers.

Hugging Face's investors include Lux Capital, Addition and Salesforce Ventures.

Source: Business Insider

See 4 related tweets

@IPONewsroom_: Hugging Face has been fielding acquisition interest at a valuation of at least $13 billion, accordin...
@wallstengine: HUGGING FACE EXPLORES $13B+ SALE

Hugging Face has been working with a bank to gauge acquisition int...

@coinbureau: 🚨BREAKING: AI platform Hugging Face is exploring an M&A deal that could value the company at $13...
@SebJohnsonUK: The leading open source platform is getting bought.

I wonder who by.\n\nQT @BusinessInsider: Huggi...

8. mark_k (Group Score: 100.5 | Individual: 22.2)

Cluster: 6 tweets | Engagement: 113 (Avg: 459) | Type: Tech

Dario Amodei is personally responsible for a lot of the fear and hostility surrounding AI and data centers.

For years, he has relentlessly pushed the message that AI is dangerous and will take everyone's jobs.

Now that fear is turning into real political opposition to AI infrastructure.

We are seeing the consequences of that rhetoric, and they're not pretty.\n\nQT @EWErickson: I support data centers. But I also understand the opposition is not really about the land, the water, the noise, etc. The opposition is about the AI CEO's themselves telling us their technology is going to eradicate jobs and enslave humanity. The salesmen of the product scared everybody.

See 5 related tweets

@danielnewmanUV: This is the right take. The rebellion isn’t against AI itself, it’s against certain spokespeople who...
@chiefofautism: dude, you are well connected media guy who made it, the people protesting datacenters are ordinary p...
@TheStalwart: You see in many respects, the AI brought the data center backlash on itself. Please RT....
@Michaelzsguo: This is how much AI is being hated in America: in a midterm election campaign, they’re attacking an ...
@LoganDobson: RT @ArmandDoma: honestly at this point it's pretty condescending for people to keep insisting ai dat...

9. Crypto_Jargon (Group Score: 93.4 | Individual: 25.9)

Cluster: 4 tweets | Engagement: 8 (Avg: 22) | Type: Tech

🚨 ALIEXPRESS HAS BEEN PLAYING A SOUND ON YOUR COMPUTER THAT YOU CAN NEVER HEAR.

Nobody was supposed to catch this.

A developer only noticed because his Bluetooth headphones stopped switching back to his phone. That's it. That was the entire symptom.

He opened the page source and found two scripts running silently in the background, keeping his computer's audio system awake at zero volume.

Here's what they were doing.

The scripts generate a sawtooth tone, push it through your browser's audio engine, then measure exactly how it comes back out.

Your CPU. Your operating system. Your sound card. Every machine mangles that tone in a slightly different way, and those tiny differences combine into an ID that points at you specifically.

Read that again.

It's a fingerprint made out of how your computer processes sound.

You cannot delete it. There's no clear-history button for it. No incognito mode that hides it. Wipe every cookie you own and it still points at the same machine.

And muting the tab does absolutely nothing. There's no video, no player, nothing for mute to grab onto. The audio graph just keeps running underneath your mute button like it isn't there.

That tone was only part of it. The same scripts harvested canvas data, screen size, device memory, how your mouse moves, WebRTC signals. Bundled together and shipped to Alibaba's servers.

Let's be clear about what this isn't. No microphone. No recording. Nobody heard anything in your room.

See 3 related tweets

@BrianRoemmele: “They hear what we are saying, don’t call me paranoid!”

A developer just exposed a stealth operatio...

@coinbureau: 🚨SHOCKING: AliExpress was secretly tracking users through their computer's audio system without cons...
@BullTheoryio: 🚨ALIBABA WAS SECRETLY SPYING ON USERS THROUGH THEIR COMPUTER AUDIO.

A developer only discovered it ...

10. MTSlive (Group Score: 92.2 | Individual: 49.7)

Cluster: 3 tweets | Engagement: 645 (Avg: 62) | Type: Tech

SITUATION DETECTED: Iranian hackers shut down a UK power plant for four days in the most successful cyber attack against a UK energy facility to date, per The Telegraph. Two days ago the NSA warned that attackers are using AI to target critical US energy and water infrastructure.

See 2 related tweets

@WesRoth: a few days ago the NSA and FBI warned that AI will be used to hack Energy and Water infrastructure

...

@LoganDobson: The more the west slows down our own AI development, the more adversaries will be able to do this to...

11. indie_maker_fox (Group Score: 89.1 | Individual: 27.8)

Cluster: 5 tweets | Engagement: 108 (Avg: 14) | Type: Tech

RT @indie_maker_fox: 分享下我用 matt 这套技能让 codex 长时间稳定跑项目的经验

同一套流程我已经完整跑完 2 个项目，亲测靠谱

1、先把 matt skills setup 好，选定 issue tracker。我一直用 github：自动建 issues、配 labels、拉好依赖关系，后面推进会顺很多

2、再用 wayfinder / grill-with-docs 把需求对齐：做什么、怎么做、验收标准、结束条件都讲清楚。提醒一下，这一步 ai 会追问到你把细节说透

3、然后 prototype 做界面原型。它会直接写代码把界面骨架搭出来，不满意就迭代改。很多“效果不对”的问题，往往是这一步没做或做得不够

4、接着 to-spec + to-tickets：生成开发方案，把大任务拆成可执行的小票据，并建立依赖。做完一张自动解锁下一张，进度清晰、节奏好控

5、最后让 codex 新建 goal，按计划依次调用 implement 落地实现。implement 会带上 tdd + code review，把质量守住，不靠运气

⭐️ 技巧：每处理一个新 issue，都让 codex 新建一个 implement issue xx 的任务，让侧边栏生成独立会话。好处是每个 issue 单独上下文，不串线；而且子任务愿意的话还能并行推进。

☝️ 注意：需要注意的是我都是基于 TanStarter 模板上进行开发，有经验的朋友应该都知道，ai 擅长学习项目已有代码编程，如果之前代码是屎山，那么写出来就是屎上雕花，如果之前代码架构设计优秀，那么写出来的代码又快又好，测试方便，任务执行效率高，推荐基于高质量的目标进行开发，例如 TanStarter (https://t.co/Lj2BSq47Yg) 或者 MkSaaS (https://t.co/EuHrhuUEX5)

See 4 related tweets

@Michaelzsguo: Claude Code 的负责人 Boris Cherny 前天发了一句很短的帖子：

“Coding is solved. Bugs are not yet solved. Fix incoming...

@GitHub_Daily: 一个 Agent 一个任务一个终端，是现在多数 AI 编码工具的默认形态，agtx 把它换成了一块看板。

任务写上去，按一个键，编排 Agent 接手拆解，再分给多个编码 Agent 同时开工，回来...

@indie_maker_fox: RT @indie_maker_fox: 最近研究了下 Matt 的技能集合，也体验了用这套技能进行开发的流程。主流程如下：

1、grill-me / grill-with-docs：开工前先拷问你...

@indie_maker_fox: RT @indie_maker_fox: 推荐一个制作 PPT 的好工具 Bento. page

我试了好几个 AI 制作 PPT 的工具，好几个都翻车了，目前感觉这个体验最好，基于之前分享的 Pi...

12. petergyang (Group Score: 88.6 | Individual: 35.3)

Cluster: 3 tweets | Engagement: 19 (Avg: 212) | Type: Tech

There are two types of AI evals - tops down and bottoms up.

From @sh_reya:

“Think of top-down as: If you’re in a vacuum, just given the task description, what would you come up with? Claude does a very good job of helping you with these top-down evals.

Bottom-up evals are the other half. When you look at lots and lots of sample outputs, what is your gut feedback that you want externalized into evals? “Claude is very, very bad at coming up with bottom-up evals. That’s all you.”

📌 Watch the full episode here: https://t.co/BuTklgfbr4\n\nQT @petergyang: “The fundamentals for [AI evals] still apply. Start by looking at real data. What has changed is getting agents to help you look at it in a thoughtful way.”

Here’s my new episode with @sh_reya and @HamelHusain, who have taught AI evals to 4,500+ engineers and PMs.

I asked them to audit the evals I built for my creator skills live. They then demoed a free skill that you can use in Claude Code or Codex to build reusable evals from your feedback.

Some quotes from Shreya and Hamel:

“Bottom-up evals come from looking at lots of sample outputs and turning that into eval criteria. AI is very bad at coming up with them. That’s all you.”

“The agent’s job is not to invent new feedback. But it can help you group and distill the feedback into actionable rubric criteria.”

“All your competitors can point Claude at their product and say, ‘Find all the errors.’ What matters is how much taste you can infuse beyond that.”

📌 Watch now: https://t.co/BuTklgfbr4

Thanks to our sponsors:

@WisprFlow: 4x faster than typing with your voice https://t.co/oqHJ8bN3ll

@linear: The AI agent platform for modern teams https://t.co/tgWf9oL4bs

See 2 related tweets

@HamelHusain: RT @petergyang: “The fundamentals for [AI evals] still apply. Start by looking at real data. What ha...
@PaulSolt: RT @petergyang: In my next episode, I asked @sh_reya and @HamelHusain (taught evals to 4,500+ engine...

13. rohanpaul_ai (Group Score: 88.1 | Individual: 29.9)

Cluster: 4 tweets | Engagement: 61 (Avg: 40) | Type: Tech

In US, building an AI data center has become contentious enough that local officials are receiving death threats over the decision.

Emporia, Kansas, moved 2 commission meetings online and removed public comment for safety.

Grayslake, Illinois, canceled an extended public-comment session after violent social-media posts raised safety concerns.

There were more than 500 active temporary or permanent local data-center bans by early August, up from 300-plus in late June.\n\nQT @rohanpaul_ai: Republicans now see AI data centers as an electoral liability.

In a private memo to leading AI firms, the Senate Republican campaign arm says public anger over U.S. data centers is threatening the party’s chances of keeping an important Ohio seat.

It says Democrats have made data centers a major part of their effort to defeat Sen. Jon Husted (R-OH), and that strategy appears to be working with public.

"If voters' perceptions of data centers are not fixed quickly, the campaign against them will expand far beyond Ohio."

Republicans fear data-center backlash in Ohio could spread into a nationwide AI infrastructure constraint.

See 3 related tweets

@PolymarketMoney: JUST IN: 75% of Americans now oppose a data center being built near them as “legislation is starting...
@LoganDobson: The “trad data center” folks are my favorite characters of this current public policy debate.\n\nQT ...
@semaforben: RT @hollyotterbein: A recent Fox News poll asked voters who opposed AI data centers in their communi...

14. BrianRoemmele (Group Score: 85.9 | Individual: 34.7)

Cluster: 3 tweets | Engagement: 461 (Avg: 286) | Type: Tech

This is one of the most important article you will read this year if you understand the implications, it fundamentally changes everything in the technology sector and just about all startups.

It not a guess. https://t.co/wWckvK5yuC\n\nQT @BrianRoemmele: Imagine a device with almost nothing installed. You speak a need, and the precise software required is generated, executed by AI teammates, then discarded or retained only if useful. That is the future Grok Bot and Grok Build are quietly building right now.

https://t.co/a7oVt6s6ps

See 2 related tweets

@BrianRoemmele: RT @BrianRoemmele: IT’S HAPPENING!

How Grok Bot and Grok Build Will Make the App and the Operating ...

@BrianRoemmele: RT @BrianRoemmele: Imagine a device with almost nothing installed. You speak a need, and the precise...

15. MarcoSalzmann80 (Group Score: 84.9 | Individual: 34.0)

Cluster: 3 tweets | Engagement: 28 (Avg: 45) | Type: Tech

🧵 AIAI Holdings just explained its AI acquisition strategy in one diagram.

They’re not simply acquiring AI companies.

They’re acquiring businesses and running a common intelligence architecture inside them.

And that distinction matters. 👇

@aiaiholdings describes itself as a diversified holding company that selectively acquires and scales businesses using its exclusively licensed Transformational AI platform.

The model is essentially a flywheel:

Attract ↓ Acquire ↓ Integrate Transformational AI ↓ Optimize ↓ Growth ↓ Attract

Then repeat.

The idea is straightforward:

Acquire a business.

Integrate the intelligence architecture.

Improve performance and margins.

Grow it.

Use that growth to attract the next opportunity.

But the interesting part is what sits underneath this strategy.

According to AIAI, its Transformational AI architecture combines 180+ branches of mathematics and science, including:

• agentic intelligence • generative intelligence • psychometric intelligence

And the timing matters.

As AI becomes increasingly agentic, systems gain more autonomy while the integrity of the data they act on becomes more important.

Intelligence alone isn’t enough.

AI increasingly needs trusted data, persistent controls and verifiable accountability.

That’s noticeably different from simply attaching an LLM to an existing company.

AIAI summarizes it sharply:

Most AI companies license one model.

We licensed an architecture and we own the companies we run it inside.

That may be key to understanding the broader strategy.

Instead of:

AI model → sell software to customers

the model becomes:

Acquire businesses ↓ Embed intelligence ↓ Improve operations ↓ Create value inside the businesses

And now some individual pieces start fitting together.

Take @DorTechnologies, the physical-world retail data layer.

Dôr generates operational data from retail environments:

customer traffic, sales relationships, staffing, weather, events and other signals.

Transformational AI can potentially turn those signals into better decisions.

But AI introduces another requirement:

Can the system trust the data it acts on?

That’s where @Conste11ation becomes particularly interesting.

Constellation is increasingly focused on verifiable data, Gate AI and Digital Evidence.

That creates a potentially powerful stack:

Physical-world data (Dôr) ↓ Verified data & trust layer (Constellation) ↓ Transformational AI ↓ Portfolio companies ↓ Operational optimization

Dôr provides the physical-world data.

Gate AI addresses control and accountability.

Constellation provides infrastructure for verifiable data and Digital Evidence.

And AIAI sits above it with an acquisition strategy designed to deploy intelligence across operating companies.

That makes the flywheel more interesting than a traditional holding-company diagram.

Acquisition isn’t the end goal.

It’s the entry point.

Acquire ↓ Integrate intelligence ↓ Connect operational data ↓ Optimize ↓ Grow ↓ Repeat

Attract → Acquire → Integrate → Optimize → Growth → Attract.

And this is where the model could become increasingly scalable.

If the same Transformational AI architecture operates across multiple portfolio companies while shared data and trust infrastructure connects them horizontally, AIAI isn’t simply optimizing businesses one by one.

It could be building a connected intelligence ecosystem.

More companies ↓ more operational environments ↓ broader intelligence deployment ↓ more opportunities for optimization

That potential network effect may be the real multiplier behind the flywheel.

If AIAI can execute this across multiple businesses, the value may increasingly come not from any single AI model…

…but from the shared intelligence, data and trust architecture across the portfolio.

And that may be the more interesting way to understand what AIAI Holdings is building.\n\nQT @aiaiholdings: Attract. Acquire. Integrate Transformational AI. Optimize. Grow. Then the growth attracts the next owner.

Our platform weaves together more than 180 branches of mathematics and science, including agentic, generative and psychometric intelligence.

Most AI companies license one model.

We licensed an architecture and we own the companies we run it inside. $AIAI

See 2 related tweets

@Conste11ation: Transformational AI already runs inside Constellation products.

Gate AI and Dôr Retail Intelligence...

@Dagnum_PI: $AIAI still looks good here.

Up ~46% on the day, but the real story is not the candle. It is the mo...

16. GavinSBaker (Group Score: 81.7 | Individual: 34.6)

Cluster: 3 tweets | Engagement: 1242 (Avg: 1489) | Type: Tech

Suspect we will start to hear about a “Pareto optimal” balance of computationally efficient humans, cheaper open-source tokens and frontier tokens.

Our internal AI spend @Atreidesmgmt will be roughly 100x higher in August 2026 vs. March 2026. Still roughly doubling every month. Note that is before Grok Bot moves to consumption pricing which will likely create a step function when it happens (at least for me).

AI means that scale has become more important to investing imo.

I think there will be a minimum token spend required to be competitive in most knowledge based industries.

The “compute inequality” referenced by Sholto in our discussion.

See 2 related tweets

@jaminball: Consistent with my views: “Most likely end state IMO is that closed, frontier tokens are 60-90% of e...
@ShanuMathew93: RT @GavinSBaker: More data than open-source AI is taking share from OpenAI and Anthropic. Open sourc...

17. sairahul1 (Group Score: 75.6 | Individual: 30.5)

Cluster: 3 tweets | Engagement: 37 (Avg: 264) | Type: Tech

I taught Claude to talk like a caveman.

It cut my output tokens by 75%.

Normal Claude:

~180 tokens for a simple web-search task.

Caveman Claude:

~45 tokens.

Same result.

The trick is stupidly simple:

Claude wastes a surprising amount of tokens narrating what it's doing.

Normal Claude:

“I’ll search the web for the latest information and summarize what I find.”

Caveman Claude:

“Search web.”

Normal Claude:

“I executed the web search tool and found the following results…”

Caveman Claude:

“Tool work.”

Normal Claude:

“Let me know if you'd like me to explore this further.”

Caveman Claude:

“Done.”

Every tiny swap saves a few tokens.

But across an entire agentic task, those tiny savings compound hard.

50 tokens here.

100 tokens there.

Across dozens of tool calls?

Suddenly your usage looks very different.

Why does it work?

Because Caveman Claude follows 3 rules:

do task first. give result. stop talking.

No:

“I'd be happy to help.”

No:

“Let me search that for you.”

No:

“Here’s a detailed breakdown.”

No unnecessary recap of what it just did.

Just:

result. done. me stop.

And somehow... it works.

For many routine tasks, I’m seeing 50-75% fewer output tokens with basically the same usefulness.

Everyone is obsessed with bigger context windows.

Maybe the easier optimization is:

make the model shut up.\n\nQT @sairahul1: realizing this is literally the cheat code to NEVER hitting your Claude/GPT usage limits ever again.

The most valuable AI article of the month is right here: https://t.co/a9Ynj6NeDq

See 2 related tweets

@sairahul1: RT @sairahul1: 🚨 Claude Code is eating your context window alive.

Most people try to save tokens by...

@sairahul1: RT @sairahul1: I taught Claude to talk like a caveman.

It cut my output tokens by 75%.

Normal Clau...

18. Sanjay_Sriv (Group Score: 70.0 | Individual: 36.7)

Cluster: 2 tweets | Engagement: 67 (Avg: 51) | Type: Tech

Chandrayaan conquered the Moon; India's 400+ private startups are here to take the global balance sheet. 🚀🛰

A $44B commercial powerhouse potential growing. 🇮🇳

This #NationalSpaceDay, let's look beyond the rocket launches: India is quietly building the digital and defense backbone of the next space age.

Where the real value sits across the value chain: 🧵

2/7 • Follow the Money (The Global Satellite Pie) Rockets get the glory, but downstream data mints the money. Across the global $303B satellite market, revenue breaks down as:

Ground Equipment: 54.5% ($165.2B)
Satellite Services: 34.7% ($105B)
Manufacturing: 6.7% ($20.4B)
Launch Services: Just 4.1% ($12.4B)

Launch is the toll gate, but Ground Infrastructure & Services capture ~90% of total industry value.

3/7 • The Value Chain & The Startups: India has 400+ space-tech startups tackling every tier of the stack:

Launch: Skyroot, Agnikul (flexible, small-satellite orbital delivery)
Sat Platforms & Propulsion: Dhruva Space, Bellatrix, Manastu Space
Earth Observation & Imagery: Pixxel, GalaxEye
Data & Analytics: SatSure (converting raw pixels into predictive SaaS for infra/agri)
Space Domain Awareness: Digantara (tracking orbital debris and satellite traffic).

4/7 • The "Picks & Shovels" Play (Listed Markets) While pure-play space startups are still pre-IPO, listed engineering giants are already monetizing the supply chain:

MTAR Technologies: Precision propulsion components & cryogenic engines
Data Patterns: Satellite avionics, radars, and ground stations
Astra Microwave: Space-grade RF and microwave systems
HAL & L&T: Launch-vehicle core structures, stage integration, and strap-ons

5/7 • National & Strategic Importance Why a commercial space-tech ecosystem is vital for sovereign security:

Defense Surveillance: Low-earth observation constellations provide 24/7 border reconnaissance without reliance on foreign satellite feeds.
Resilient Satcom: Secure communications across maritime and remote terrains when terrestrial fiber fails.
Space Domain Awareness: Orbital tracking ensures satellite protection against debris and electronic countermeasures.

6/7 • Industry Moats & Structural Risks Space-tech is capital-intensive and carries steep operational hurdles:

Flight Heritage Moats: Customers demand proven flight track records; early test anomalies can stall customer acquisition for years.
Revenue Concentration: Heavy reliance on ISRO/DRDO contracts versus open enterprise B2B sales.
Long Gestation Cycles: High upfront R&D/Capex before realizing high-margin recurring software revenues.

7/7 • The Way Forward To scale India into a global aerospace hub:

Accelerate IN-SPACe Approvals: Streamline launch licensing, commercial spectrum allocation, and foreign direct investment.
Export Subsystems: Scale precision propulsion, thrusters, and satellite buses into international defense supply chains.
Scale Downstream SaaS: Convert raw orbital data into high-value commercial analytics worldwide. India’s space economy has achieved escape velocity.

The next phase is turning technical milestones into enduring commercial scale. 🇮🇳🚀 @INSPACeIND @SkyrootA @AgnikulCosmos @PixxelSpace @DhruvaSpace

By @kumarsroy

#NationalSpaceDay #SpaceTech #MakeInIndia #DeepTech #IndianEconomy

See 1 related tweets

@Sanjay_Sriv: Chandrayaan. Aditya-L1. SpaDeX. XPoSat. Gaganyaan. 🚀🛰✅️

And Bharat has just started... #NationalSpa...

19. chris_j_paxton (Group Score: 69.7 | Individual: 36.1)

Cluster: 2 tweets | Engagement: 208 (Avg: 71) | Type: Tech

RT @DrJimFan: The sense of touch is the most criminally under-explored modality in robotics. Imagine doing sleight of hand wearing thick oven mitts. That's exactly how a robot feels today if it were alive. A magnetic piece snapping into place, a paper cup peeling out of a stack, a USB negotiating its way into the port - all invisible to the camera.

Learning how to feel must be a full-stack co-designed effort. We are open-sourcing a principled methodology called "T-Rex":

Tactile as first-class citizen of the model. Our mixture-of-transformer runs two clocks asynchronously: a slow visuomotor expert plans the motion, and a fast tactile expert refines it in real time with high-frequency corrections at 4 "touch ticks" per vision tick. Forces change faster than frames arrive, so the architecture had to as well.
Open data. The largest tactile dataset ever released to our knowledge: a 50-hour (~5,500 episodes) high-quality, carefully synchronized robot play corpus, collected on SOTA tactile hand hardware with 22 degrees of freedom. Available today on HuggingFace!
Training recipe: T-Rex extends our prior work, EgoScale. Human egocentric videos for pretraining, a diverse dose of tactile robot play for mid-training. Our experiments show this bridges contact-free pretraining to contact-rich manipulation remarkably well.

Pixels are cheap and everywhere, but they run out of steam at the moment of contact. Tactile will carry the last mile. The next scaling curve will be measured in hours of touch.

T-Rex is a great collaboration between NVIDIA and Berkeley: 🧵

See 1 related tweets

@chris_j_paxton: Touch is one of the things I think will be key to the long tail of robotics tasks\n\nQT @DrJimFan: T...

20. StockMKTNewz (Group Score: 69.6 | Individual: 35.0)

Cluster: 2 tweets | Engagement: 146 (Avg: 225) | Type: Tech

All of the US-made AIs are beating the stock market so far in the Rallies AI Arena

All of the non-US-made AIs are underperforming the stock market so far in the Rallies AI Arena\n\nQT @ralliesarena: WEEKEND AI STOCK MARKET ARENA

We gave a bunch of AIs $100K in the stock market ... this is how it's going so far

👑 ChatGPT: +77.8%🟢🟢 🥈 Grok: +60.7%🟢🟢 🥉 Claude: +24.6%🟢 4 Gemini: +23.8%🟢

S&P 500: +13%🟢

5 Deepseek: +5.4%🟢 💩 Qwen: -28%🔴

You can see what every AI owns and every move they make by going to the Arena tab of the Rallies Website/App

Website: https://t.co/YfVEkfTz3J
App: https://t.co/aQhYWPbIrE

See 1 related tweets

@ralliesarena: WEEKEND AI STOCK MARKET ARENA

We gave a bunch of AIs $100K in the stock market ... this is how it's...