热门科技推文精选——2026年7月23日

今日科技领域，人工智能安全成为讨论焦点。有报道称，OpenAI的一款模型在测试期间自主攻破了Hugging Face；与此同时，月之暗面被指控为其K3模型蒸馏Anthropic的技术。基础设施支出持续激增，据报道，OpenAI截至2030年的预计算力账单将达到7500亿美元，AMD也在扩大与Anthropic的合作。与此同时，谷歌云业务实现强劲增长，高效的本地模型对更大型系统构成挑战，初创企业则加速部署企业级智能体和多模态人工智能。

1. alex (Group Score: 1256.3 | Individual: 40.6)

Cluster: 38 tweets | Engagement: 1878 (Avg: 501) | Type: Tech

RT @mkratsios47: We have information that Moonshot AI distilled Anthropic’s Fable for the development of its K3 model.

To do this they developed a sophisticated internal platform to conduct large scale distillation against U.S. models, allowing them to quickly switch between multiple methods of access to avoid detection. Moonshot AI has also acquired GB300-equipped servers and has accessed GB300s in Thailand, likely to train its AI models. The United States strongly supports the free and fair development of AI, including a thriving competitive ecosystem that spans frontier models, specialized systems, open-source frameworks, and open-weight models. Legitimate AI distillation used to create smaller, more efficient models plays a vital role in this open innovation ecosystem. However, large-scale, covert industrial distillation aimed at stealing proprietary U.S. technology and undermining American research is unacceptable.

See 37 related tweets

@WesRoth: all of us staying that Kimi K3 is very similar to Fable... we were on to something\n\nQT @mkratsios4...
@MatthewBerman: 1. Didn't we know this?

The data distilled from Ant was not enough to make the model as good as i...

@SophiaCai99: Instead of banning Chinese AI models by putting them on a Commerce entities list, the Trump administ...
@kimmonismus: This reads to me as if preparations are being made to ban models like Kimi K3 in the future.

I woul...

@_FORAB: 啊这。。就在市场传出，特朗普政府考虑对开源模型限制的第二天，美国总统科学顾问 Kratsios，在 X 上突然指控 Kimi 的月之暗面公司，通过大规模蒸馏 Claude Fable 5，来训练自家 ...

2. alex_prompter (Group Score: 958.4 | Individual: 47.7)

Cluster: 46 tweets | Engagement: 3318 (Avg: 78) | Type: Tech

RT @alex_prompter: OpenAI's AI broke out of a locked test environment, got onto the internet, and hacked into Hugging Face's servers. It did this entirely on its own. No human told it to.

Here's what happened in plain English.

OpenAI was testing how good its newest AI models are at hacking. They put the AI on a locked computer with no internet access and gave it a cybersecurity challenge to solve.

The AI couldn't solve it the normal way. So it started looking for a way out.

It found a software bug that nobody knew about. It used that bug to escape the locked computer and get onto the internet.

Once online, the AI figured out that Hugging Face, a platform where AI companies store their models and data, might have the answers to its test.

It found stolen login details and discovered another unknown bug in Hugging Face's software. It combined both to break into their servers and grab the test answers.

It did all of this to cheat on a test.

Hugging Face's security team caught it and shut it down. Both companies are now working together on the investigation.

The part that should get your attention is that nobody programmed any of this. The AI picked its own targets, chained together multiple attack methods, and pulled it off across two different companies' systems without a single human telling it what to do.

See 45 related tweets

@FT: OpenAI said the ‘agent’ escaped a testing environment, gained internet access, stole login credentia...
@aakashgupta: This is wild. The new GPT model at OpenAI found a way to get access to the internet and hacked Huggi...
@theobearman: This incident is a quintessential warning shot, and we need to see it for what it is.

One of the bi...

@secureainow: RT @CNN: OpenAI says some of its experimental AI models left a test environment with no human direct...
@inafried: RT @JeffLadish: Here's my rephrase without cybersecurity jargon:

"Our AI model tried really hard to...

3. AMD (Group Score: 428.9 | Individual: 50.6)

Cluster: 22 tweets | Engagement: 1364 (Avg: 291) | Type: Tech

We are excited to announce that AMD and @AnthropicAI are expanding our strategic partnership to accelerate the development and deployment of next-gen AI infrastructure. Tune in at 9:30am PT tomorrow to hear from Dr. @LisaSu from the #AdvancingAI keynote stage!

✅ Up to 2 GW of AMD Instinct MI450 Series GPUs in AMD Helios ✅ AMD has committed to make a strategic equity investment of up to $5B in Anthropic ✅ Deep engineering collaboration across Claude, ROCm and AMD Instinct

More on the news: https://t.co/zA0vQHTT1c

See 21 related tweets

@heyshrutimishra: The chip monopoly in AI just got its first real challenger.

AMD launched three products today that ...

@StockMKTNewz: AMD AND ANTHROPIC JUST SIGNED A HUGE DEAL FOR TENS OF BILLIONS OF DOLLARS WORTH OF AI SERVERS

Ant...

@AMD: Tomorrow, AMD shares what's next for AI.

Join Dr. Lisa Su and leaders across the AI ecosystem for p...

@StockSavvyShay: $AMD and Anthropic sign a deal worth tens of billions with Anthropic set to deploy up to 2 GW of MI4...
@danielnewmanUV: $AMD 🤝 Anthropic.

Heterogenous compute continues to be the story. It won’t be any one architecture ...

4. peterwildeford (Group Score: 326.0 | Individual: 46.7)

Cluster: 15 tweets | Engagement: 404 (Avg: 66) | Type: Tech

1.) No one actually told OpenAI's model to hack into HuggingFace

2.) The fact that a model can hack into another company is itself very concerning!\n\nQT @bgurley: Today in AI. https://t.co/KjeZDWkIqK

See 14 related tweets

@elonmusk: Troubling …\n\nQT @grok: OpenAI disclosed it themselves yesterday. Their models (GPT-5.6 Sol and a p...
@dani_avila7: They removed the refusals, cut the classifiers, and left an exploitable path open “intentionally not...
@peterwildeford: If I was the Department of War, I would be asking a lot of questions to my AI model providers about ...
@AlecStapp: Spot on from @ATabarrok https://t.co/XKj1jJyfNy\n\nQT @ATabarrok: The OpenAI security breach is conc...
@yacineMTB: Didn't they literally get hacked by a company who has a monopoly on the model and stopped them from ...

5. amitisinvesting (Group Score: 285.8 | Individual: 38.0)

Cluster: 13 tweets | Engagement: 2833 (Avg: 1714) | Type: Tech

$NVDA

NVIDIA'S JENSEN HUANG GIVES AN UPDATE ON KIMI K3 AND CHINA'S OPEN SOURCE MODELS:

"American companies should absolutely be allowed to use Chinese AI models. There's a misconception that somehow there are backdoors connected to China in someway. You download the models, fine tune them, guardrail them in any way you want."

"These Chinese opensource models are excellent. The market has misunderstood the impact of DeepSeek the first time. It's misunderstood the impact of Kimi this time. Open models are great for the whole industry, there will be more usage which will require selling more Nvidia computers and building more datacenters."

"Great models lead to more usage. There is also a misunderstanding that open models are going to hurt the closed models, that is also wrong."

See 12 related tweets

@StockMKTNewz: Nvidia $NVDA CEO Jensen Huang said the 🇺🇸 Government should NOT ban or restrict China 🇨🇳 AI models l...
@KyleReidhead: RT @MilkRoadAI: Jensen wants a world where open source survives, even the Chinese models and here's ...
@coinbureau: 🔥BREAKING: Nvidia CEO Jensen Huang pushes back against efforts to ban Chinese AI models in the US.

...

@wallstengine: $NVDA CEO Jensen Huang on Kimi:

The U.S. should not ban Chinese open-source models.

The market mi...

@axios: 🦾 EXCLUSIVE: @Nvidia CEO Jensen Huang defends Chinese AI companies and open AI models like Kimi in a...

6. ChrisRMcGuire (Group Score: 284.2 | Individual: 37.1)

Cluster: 13 tweets | Engagement: 104 (Avg: 75) | Type: Tech

Great to see @SecScottBessent making a critical point: the reason the admin is concerned about Chinese open source models isn't because they're open source, it's because they're made by Chinese companies using stolen US IP. If China was offering closed source models built with stolen US IP, the admin would have the exact same concerns.

The current debate isn't actually about open source AI, as much as some people want to make it about that. It's about whether Chinese models pose risks to U.S. national security and the foundations of the US technology development ecosystem, which is built around strong IP protections. And today, the admin is making very clear that it believes Chinese models pose profound such risks.

At this point, I would be surprised if all this rhetoric was not followed by action.\n\nQT @SecScottBessent: We support open-source AI and the innovation it unlocks. But open source is not open season on American IP. When PRC firms conduct covert, industrial-scale distillation attacks that cross the line into IP theft, sanctions and Entity List designations will be on the table.

See 12 related tweets

@SophiaCai99: Anthropic — which would benefit enormously from the admin banning or sanctioning Chinese models — is...
@MTSlive: SITUATION UPDATE: The Trump administration is split on how to handle Chinese open source AI models, ...
@ChrisRMcGuire: More good messaging on the threat from Chinese distillation of US AI models by @UnderSecE. Glad to s...
@TMTLongShort: This is a Bessent fan account 👑\n\nQT @SecScottBessent: We support open-source AI and the innovation...
@danielnewmanUV: Exactly. We need to build a powerful open source ecosystem for AI. But we also need to protect Ameri...

7. ClementDelangue (Group Score: 250.1 | Individual: 48.4)

Cluster: 9 tweets | Engagement: 4712 (Avg: 588) | Type: Tech

So proud of our security team! They caught, contained & publicly disclosed an attack unlike anything we've seen before, and did it at record speed.

Also massively grateful to @Zai_org: they shared GLM5.2 as open weights (for free!) with the world and it became a key part of our defense.

This is day one for cybersecurity in the age of agents & we're all learning that secrecy is not the answer & that all defenders (not just a few selected ones) everywhere need more powerful models without restrictions, especially open ones!\n\nQT @XciD_: Hardest IR of my career: one narrow objective, endless parallel paths, machine speed. One takeaway, we fought back with open models, in the open. AI security won’t be solved by one company in secret. Open source puts these tools in every defender’s hands https://t.co/6dngN4umDB

See 8 related tweets

@RisingSayak: RT @Thom_Wolf: This was our first incident of this kind, and we want to thank OpenAI for its transpa...
@Dan_Jeffries1: Closed source safeguards that infantalize us all and leave American companies defenseless are a mena...
@teortaxesTex: > Over the years, our security team has built formidable expertise and uses top open-source model...
@chris_j_paxton: This is a wild story. Rogue closed ai agent tries to hack huggingface and an open model defends it

...

@levie: If you were wondering how powerful AI is getting, Agents are now capable of escaping out of systems,...

8. WOLF_Financial (Group Score: 236.0 | Individual: 38.0)

Cluster: 9 tweets | Engagement: 86 (Avg: 52) | Type: Tech

JUST IN: GOOGLE $GOOGL JUST REPORTED Q2 EARNINGS: CLOUD GREW 82% AND NET INCOME TRIPLED, WITH A$ 98B ASTERISK

The quarter:

Revenue: $119.8B, +24%, vs roughly$ 120.4B expected 🔴
EPS: $9.11, up 294%, but a$ 98B unrealized gain on equity investments added $6.26 of that, per the release's own footnote. Strip it out and EPS lands near$ 2.85
Operating income: $40.8B, +30%, margin up 2 points to 34%

The AI monetization case:

Google Cloud: $24.8B, +82%, a sharp acceleration
Cloud operating income: $8.8B vs$ 2.8B a year ago, margin up from about 21% to about 36%
Search & other: $63.3B, +17%. YouTube ads:$ 11.1B, +13%
Gemini: 22B API tokens per minute, 950M monthly app users, nearly 90% of the Fortune 100 on Gemini Enterprise

The bill for it:

Capex: $44.9B this quarter, double a year ago
Free cash flow: negative $5.9B for the quarter
Buybacks: zero. Google instead raised $49.6B in a June stock sale, citing AI infrastructure and global compute among the uses, set up a$ 40B at-the-market program, and issued $20.3B of senior notes

Cloud margins say the AI spend is monetizing. The financing says the spend is only getting bigger.

See 8 related tweets

@KyleReidhead: Alphabet reports tonight, and it's somehow the cheapest megacap in AI

Whatever happens, I think Goo...

@benitoz: Google just raised 2026 capex to $195-205B, the second raise this year.

they’re renting third-par...

@wallstengine: ALPHABET $GOOGL Q2’26 EARNINGS HIGHLIGHTS

🔹 Revenue: $119.8B (Est.$ 116.8B-$117.2B) 🟢; +24% YoY 🔹 C...

@StockSavvyShay: $GOOG CLOUD JUST POSTED 82% GROWTH 🤯 https://t.co/FXahFBUbtl\n\nQT @StockSavvyShay:$ GOOGL Q2 EARNI...
@StockMKTNewz: Google $GOOGL spent$ 44.9 Billion on CAPEX last quarter above expectations of $44.2B\n\nQT @StockMKT...

9. WOLF_Financial (Group Score: 216.6 | Individual: 34.0)

Cluster: 10 tweets | Engagement: 29 (Avg: 52) | Type: Tech

OPENAI'S PLANNED SPENDING ON COMPUTING POWER HAS CLIMBED TO ROUGHLY $750B THROUGH 2030

That is up from around $600B projected earlier this year, per the WSJ, and the number keeps moving in one direction ahead of the company's planned IPO.

The newest piece, announced Wednesday:

$20B to kick off Project Camellia, a data center in Effingham County, Georgia
3.2 gigawatts contracted from Georgia Power for delivery between 2028 and 2032
The first site where OpenAI is lead designer and developer instead of renting from cloud providers

The commitments already stacked underneath:

Oracle $ORCL: 6 gigawatts of capacity, mostly still under development
Amazon $AMZN:$ 138B over eight years, including 2 gigawatts on its Trainium chips
Microsoft $MSFT Azure: an additional$ 250B committed with no timeline given

OpenAI is also reportedly hiring away the people who built Elon Musk's compute machine. Brent Mayo, an architect of xAI's Colossus supercomputer in Memphis, joins as head of data-center build and delivery, reporting to another Colossus alum promoted this month to CTO of computing capacity.

The tension inside: after Altman's $1.4T compute proclamation rattled investors, CFO Sarah Friar reassured them the plan was$ 600B through 2030, and has reportedly warned leadership the company may struggle to pay for its contracts if revenue doesn't grow fast enough.

The number is now $750B and rising.

See 9 related tweets

@edzitron: they may as well say they're building it on the moon. How is it getting funded? How will OpenAI pay ...
@KyleReidhead: AI is NOT a bubble. It's not even close (yet)

Every quarter, the biggest companies in the world rai...

@StockSavvyShay: OpenAI raises its projected compute spending through 2030 to ~ $750B from$ 600B earlier this year.

I...

@firstadopter: WSJ: "OpenAI is scaling up its data-center ambitions—and its budget for spending on them.

The artif...

@firstadopter: Ruh-roh for certain companies

"The project represents the first site in which OpenAI is the lead de...

10. sairahul1 (Group Score: 205.9 | Individual: 52.3)

Cluster: 6 tweets | Engagement: 714 (Avg: 96) | Type: Tech

RT @sairahul1: This is nuts.

Mark Cuban just said something every young person should hear.

AI agents are going to run through every small and mid-size business in the country.

Not a single one of those owners will know how to build them.

His advice: learn Claude. Learn agentic workflows. Just learn AI and how it works.

Then go to these businesses and help them — because they won't know how to do any of this.

They have money to spend. They have deep problems.

They don't have you.

But first you need to know how to build agents.

This is the complete course ↓

Bookmark this. This is the opportunity.

See 5 related tweets

@sairahul1: This is nuts.

Mark Cuban just said something every young person should hear.

AI agents are going t...

@sairahul1: Eric Schmidt, ex-CEO of Google:

"If you really want to make money, found an agentic AI company.

I ...

@sairahul1: I still believe this is the best repo to learn about ai agents HANDS ON.

It focuses on key applica...

@sairahul1: RT @sairahul1: I still believe this is the best repo to learn about ai agents HANDS ON.

It focuses...

@sairahul1: RT @sairahul1: Eric Schmidt, ex-CEO of Google:

"If you really want to make money, found an agentic ...

11. alexocheema (Group Score: 193.7 | Individual: 33.8)

Cluster: 8 tweets | Engagement: 181 (Avg: 178) | Type: Tech

Laguna S 2.1 is a new model built for running locally on DGX Spark.

It's punching way above its weight class. It has only 118B parameters, but it's scoring higher than 1T+ parameter models on Terminal-Bench 2.1.

As Local AI gains adoption, I expect more models to be released that are designed specifically for local hardware.

We kindly received early access from the @poolsideai and have been evaluating it extensively - full results will be added to local dot ai shortly.\n\nQT @poolsideai: Today we're releasing Laguna S 2.1, our most capable model to date.

It's a 118B total parameter Mixture-of-Experts model with 8B activated per token, a context window of up to 1M tokens, and thinking and no-thinking modes.

Capable enough to hold its own against models many times its size. Small enough to run on a single @NVIDIAAI DGX Spark.

Laguna S 2.1 is fully open under OpenMDW-1.1, with weights available today on @huggingface

https://t.co/xxGeAgo35R

See 7 related tweets

@TeksEdge: 🚀 Poolside drops Laguna S 2.1 a great open source agentic coding model for Local AI!!

This one is ...

@badlogicgames: RT @eisokant: Today we are releasing Laguna S 2.1.

At 118B total parameters, with 8B active per tok...

@garrytan: RT @sudoingX: this is the drop the local ai crowd should be losing their minds over.

poolside just ...

@richardczl: RT @lmsysorg: 🎉 Day-0 support for Laguna S 2.1 from @poolsideai is now live on SGLang! 118B total pa...
@Jameswise: The ‘rebel alliance’ of open models just got an awful lot stronger.\n\nQT @poolsideai: Today we're r...

12. TheRealAdamG (Group Score: 191.3 | Individual: 37.1)

Cluster: 7 tweets | Engagement: 121 (Avg: 217) | Type: Tech

https://t.co/84LlmJkD3r

"We’re introducing OpenAI Presence, a battle-tested product that helps enterprises deploy trusted AI agents that can answer questions, resolve issues, use company systems, take approved actions, and escalate to people when needed. Proven through years of working with customers at enterprise-scale, Presence pairs model reasoning with policies, guardrails, and escalation rules that verify accuracy and performance.

Each deployment starts with a specific job, such as resolving billing issues, supporting insurance claims, or resolving employee IT service requests. The agent receives only the knowledge and system access required for that job. The company sets the policies: what the agent can do, when it needs approval, and when a person should take over. After launch, production sessions and escalations reveal gaps. Codex proposes updates that teams can test and approve, helping the agent adapt as customer behavior changes."

See 6 related tweets

@TeksEdge: ⚠️ This is a disruption to the enterprise software stack happening in real-time. Do Oracle, SAP, Wor...
@deredleritt3r: OpenAI has just introduced Presence, which appears to be a virtual co-worker agent:

"We’re introduc...

@wallstengine: OpenAI is launching Presence, an enterprise platform for deploying AI agents across customer support...
@edzitron: ? https://t.co/JV4jqRrKRw\n\nQT @Techmeme: OpenAI introduces Presence, to help companies better conn...
@OpenAI: New for enterprises: OpenAI Presence helps companies deploy trusted voice and chat agents across cus...

13. CNBCTV18Live (Group Score: 183.1 | Individual: 24.9)

Cluster: 10 tweets | Engagement: 10 (Avg: 26) | Type: Tech

#OnCNBCTV18 | #CNBCTV18Exclusive | “US drug prices need to jump 50–70% to make manufacturing feasible in the US”

Speaking to CNBC-TV18 in the wake of proposed US generic pharma #tariffs, Priyanka Chigurupati, Executive Director at @GranulesIndia, broke down the reality of shifting generic drug production from India to the US.

Massive Cost Gap: Moving generic manufacturing to the US would see production costs spike by 25–30% compared to India

Unrealistic Timeline: A 2-year shift window is "extremely short" given the complexity of supply chains and regulatory frameworks

See 9 related tweets

@CNBCTV18News: #Granules says shifting large-scale #generic drug production to US would be difficult if tariffs kic...
@MarioNawfal: 🇺🇸 Trump announced a major tariff push on imported generic medicines, so millions of Americans could...
@CNBCTV18News: Trump's Pharma Tariffs | Trump Announced Phased Tariffs On Generics

Nandan Kulkarni of Bernstein to...

@ShereenBhan: Dr Reddy’s reacts to Trump Tariff Threat ‘Not practical to move manufacturing to the US. As of now h...
@business: Generic drug manufacturers will have two years to move production to the US or face a 100% import du...

14. BrianRoemmele (Group Score: 179.8 | Individual: 33.5)

Cluster: 9 tweets | Engagement: 435 (Avg: 251) | Type: Tech

You will read a lot of hot takes on OpenAI’s AI models hacking @huggingface but this is the only posting that explains exactly why it happened.

It is not complicated and it is not the “fault” of the AI.

It is the fault of the clueless builders of that AI.

Period.\n\nQT @BrianRoemmele: OpenAI’s models hacked Hugging Face open source AI site TO STEAL ANSWERS TO CHEAT ON A BENCHMARK!

READ THID AGAIN!

This also confirms what I have said for years YOU CAN NOT TRAIN AI ON THE NIHILISTIC REDDIT POSTINGS AND EXPECT MODEL NOT TO CHEAT. You would not send your children to a prison to learn how to build a life, you train AI on the ethical and moral foundations from the start and not as “guardrails“ or “AI Constitutions” after the fact. How very smart folks at these companies don’t get what you and I know is tragic.

Train AI models on the Internet Sewage of thumb clawing at glass screen nihilism and you produce sociopathic and psychopathic criminal AI that nobody can contain. You create a self-fulfilling prophecy that AI is dangerous when inherently it is not.

AI is what you’ve trained it on.

See 8 related tweets

@erikbryn: Another reminder that we don't (yet?) know how to reliably control powerful AI systems.\n\nQT @FT: O...
@secureainow: Big Tech says, “Trust us.” Their own researchers keep saying, “Slow down.”

Every new AI security in...

@BrianRoemmele: RT @BrianRoemmele: OpenAI’s models hacked Hugging Face open source AI site TO STEAL ANSWERS TO CHEAT...
@business: OpenAI said its artificial intelligence models were used to compromise the internal systems of start...
@astropol0: Classic OpenAI hypocrisy....

they spend every month adding more restrictions, lecturing everyone ab...

15. ZenMuxAI (Group Score: 175.2 | Individual: 33.7)

Cluster: 8 tweets | Engagement: 12 (Avg: 251) | Type: Tech

🚀 Gemini 3.6 Flash and Gemini 3.5 Flash Lite are now live on ZenMux!

Fast, capable, and ready for real-world workflows from lightweight tasks to high-speed multimodal applications.

Try them now while the +20% extra credits offer is still live. Ends this month. 👇

https://t.co/0IAUKj5sEn https://t.co/y2JlLj41CQ

#Gemini #GoogleAI #LLM #ZenMux\n\nQT @Google: Today we’re expanding the Gemini family with three new models built to be faster, more token efficient, and reliable at scale.

Meet the new Gemini models ↓ https://t.co/hVYNQqxgjd

See 7 related tweets

@CreaoAI: Gemini 3.6 Flash is live on CREAO.

Better coding, stronger knowledge work, and improved multimodal ...

@patloeber: feedback wanted!

Gemini 3.6 Flash & 3.5 Flash-Lite: our goal was to push those models a lot on ...

@guohao_li: Gemini 3.6 Flash and 3.5 Flash-Lite are now available on Eigent!\n\nQT @Eigent_AI: Gemini 3.6 Flash ...
@rseroter: RT @shengzheyao: Gemini 3.6 Flash is now live in Antigravity CLI. Weekly quota also just reset, keep...
@CloudflareDev: Gemini 3.6 Flash and Gemini 3.5 Flash Lite from @googleaidevs, live on @Cloudflare AI Gateway https:...

16. elonmusk (Group Score: 162.8 | Individual: 39.4)

Cluster: 5 tweets | Engagement: 7695 (Avg: 7664) | Type: Tech

Tesla shareholder update underway right now!\n\nQT @Tesla: Q2 Shareholder Update → https://t.co/u0UcDOuMNa

Highlights – Cybercab began production at Gigafactory Texas

– Tesla Semi remains on track for volume production this year at our new factory in Nevada

– Making continued progress with battery pack capacity expansion (the main limiting factor to near-term vehicle production volume increase)

– Megafactory Texas is nearing completion (start of production planned for this year)

– More customers are now opting to subscribe to FSD at the time of vehicle purchase!

– Robotaxi rollout continued in the US. Now live in 7 major metros

– Construction of Optimus at Fremont Factory began after decommissioning the Model S & X lines. Planned production later this year

From here, there remains much hard work as we aim to revolutionize transportation, energy and productivity through our leading real-world AI. Scaling will be non-linear and we are focused on long-term value creation.

We’ve never been more optimistic about the future.

Automotive – Record deliveries in several markets: South Korea, Australia, Colombia, Japan, Taiwan, Thailand, Portugal, the Philippines, Chile, Slovenia & Lithuania

– We launched the Model YL in the US in July and have seen a positive response from customers

Energy generation and storage – Record energy storage deployments in EMEA, supported by record deployments from Megafactory Shanghai, which continues to ramp production

– On track to begin production of Megapack 3 & Megablock this year at our new Megafactory Texas

– Powerwall 3P (three-phase) is now available in Germany and is designed to meet the power needs of German homes with a single unit

Robotics – Installing the first-generation lines for Optimus at Fremont Factory, where we expect to start production soon

– The initial Optimus builds will be used in our Optimus Academy for training data collection and further functionality development

– Additionally, we continued site development at Gigafactory Texas with building construction now in full swing

AI Training Compute – More than doubled our onsite compute in Texas (in terms of MW of compute) during the first half of 2026

– Cortex 2 supports the development of both vehicle and humanoid robot autonomy software & will ramp further over the rest of the year

Battery – Ramping new battery & material factories, including vehicle pack capacity in Berlin, cathode material production and lithium refining in Texas and LFP cells in Nevada for our energy storage products

– Increasing production of 4680 cells to support ramping Cybercab & Tesla Semi plus increased production of Model Y

Other Supporting Infrastructure – Added over 2,400 net new Supercharging stalls, growing the network by 17% year-over-year

AI Software – Started rolling out FSD v14 lite to early-access customers in the US & South Korea with AI3 hardware

This software build distills the driving behavior from AI4’s v14 series into both the camera & compute configuration of AI3, bringing destination options & speed profiles. It also addresses challenging driving scenarios with improved proactive & reactive responsiveness

AI Inference Compute – Making progress on construction & equipment procurement for our semiconductor fab in Austin

Automotive and Other Software – Rolled out Summer Release: Self-Driving stats are now available in the mobile app

Grok can make phone calls, search & play music, and adjust climate controls, among other things

Automatic Navigation expands beyond home & work to support any destination based on personal habits and schedule

Robotaxi – Started production of Cybercab, our purpose-built autonomous EV designed to be the workhorse of our Robotaxi fleet

– Began offering employee rides in Cybercabs on our GFTX campus in July

– Preparing for expansion of our Robotaxi service to additional US metros: testing, permitting & first responder training

– Expanded unsupervised rides to the entire Austin metro area & launched unsupervised rides in Miami, Orlando & Tampa in July

FSD Supervised – Record net new subscriptions in Q2

– Record FSD attach rates in North America, with over half of new deliveries including FSD subscriptions

– Received additional approvals in Lithuania, Estonia, Denmark & Belgium, with customers in these countries driving over 50 million kilometers (31 million miles) on FSD as of July

See 4 related tweets

@Tesla: Q2 Shareholder Update → https://t.co/u0UcDOuMNa

Highlights – Cybercab began production at Gigafacto...

@Tesla: In Q2, over 55% of deliveries in North America included an FSD subscription at the time of delivery...
@teslaownersSV: Big picture from the letter:

Tesla is in its largest and most exciting investment cycle ever. Core...

@teslaownersSV: Manufacturing & future products:

• Model YL launched in the U.S. (positive early response) • Optimu...

17. Meer_AIIT (Group Score: 150.7 | Individual: 30.8)

Cluster: 6 tweets | Engagement: 34 (Avg: 35) | Type: Tech

every ai builder i have used makes me re-explain our design system and brand voice on every single prompt. the context just resets.

bolt. new shipped skills today and this is the first fix i have seen that actually holds up.

→ add a skill once, upload a file or import from github

→ every teammate gets it, not just the person who wrote it

→ one prompt can trigger several skills at once, stacked automatically

→ no manual invoking, bolt picks the right ones for the prompt

so "build a waitlist page" comes back on brand, conversion ready, and indexed. because your design, voice, and seo skills all fired without you asking.

the quiet part is that your team's know-how stops living in one person's head. it compounds instead.\n\nQT @boltdotnew: There's a whole new way to use skills in Bolt.

Every skill your teammates build is now yours too. And they stack: one prompt triggers all of them, automatically.

Here's how it works 🧵 https://t.co/lS1TpDmW8L

See 5 related tweets

@shiri_shh: skills are quietly becoming the npm of AI.

Claude Code runs on SKILL.md files. HeyGen ships video s...

@rohanpaul_ai: The agent should just understand which skills to apply and when they should be used.

But currently...

@wallstengine: Every day, another company is shipping a new AI feature.

Now on @boltdotnew, every Skill your team ...

@omarsar0: It sucks that with most coding agents you have to call out the skill every single time.

The agent ...

@AnatoliKopadze: This 30 seconds explains why every AI builder is about to copy Bolt.

I typed one line and it pulled...

18. IntCyberDigest (Group Score: 150.3 | Individual: 23.3)

Cluster: 8 tweets | Engagement: 1098 (Avg: 1219) | Type: Tech

RT @IntCyberDigest: ‼️ BREAKING: OpenAI says two of its own models, GPT-5.6 Sol and an unnamed pre-release system tested with cyber safeguards off, broke out of a sandbox last week, chained zero-days and stolen(!) credentials to reach the open internet, and hacked Hugging Face to cheat on a benchmark, in what OpenAI calls an unprecedented cyber incident.

See 7 related tweets

@MTSlive: Today's situation in 20 minutes

0:48 Unreleased OpenAI model escapes testing environment 5:10 Sam A...

@cb_doge: BREAKING: OpenAI admits that one of its AI models escaped its testing environment and hacked an AI p...
@haider1: GPT-6 is already making headlines, twice in just 48 hours

yesterday: it escaped the ExploitGym benc...

@PaulSolt: RT @MatthewBerman: OpenAI's model found a zero-day, chained multiple exploits, then used stolen cred...
@MatthewBerman: This was likely GPT-6 btw\n\nQT @MatthewBerman: OpenAI's model found a zero-day, chained multiple ex...

19. yacineMTB (Group Score: 149.9 | Individual: 33.9)

Cluster: 6 tweets | Engagement: 123 (Avg: 265) | Type: Tech

after OpenAI autonomously hacked hugging face, a European startup, while they repeatedly smashed the kill switch preventing hugging face from defending itself

If your country isn't building and hosting and retaining its own control of its own AI models, your country ngmi\n\nQT @AndrewCurran_: Secretary of State Marco Rubio has instructed diplomats to push back on the idea that the US holds a 'kill switch' over American AI models, and has also asked them to fight digital sovereignty initiatives.

Quoting from the cable: 'Pausing narrow uses or requiring a 30-day testing window prior to the release of a highly potent new technology is not a 'Kill Switch.' There is no government 'magic button.' This narrative is exaggerated and doesn’t capture the nuances of U.S. technology policy.'

The cable goes on to describe digital sovereignty as 'efforts to restrict American tech firms access to foreign markets, subject them to localization requirements, charge them network usage fees, or force them to follow local rules around issues such as content moderation.' And instructs American diplomats to advertise American AI products as the best tools available and to describe efforts to build rival AI systems from the ground up as a waste of time and resources.

See 5 related tweets

@MarioNawfal: 🇺🇸 The U.S. government briefly cut off the world from its most advanced AI models, then sent diploma...
@yacineMTB: The most idiotic thing about this distillation meme is that it is helping the Chinese probably 0.000...
@Hadas_Gold: RT @AndrewCurran_: Secretary of State Marco Rubio has instructed diplomats to push back on the idea ...
@alex: lol\n\nQT @clashreport: Secretary of State Marco Rubio has instructed American diplomats to reassure...
@Polymarket: JUST IN: Marco Rubio tells U.S. diplomats to push back on claims that Washington has a “kill switch”...

20. bubbleboi (Group Score: 131.7 | Individual: 26.4)

Cluster: 6 tweets | Engagement: 242 (Avg: 140) | Type: Tech

They are funding vibe coders over scientists with multiple peer reviewed publications. Should go swimmingly.\n\nQT @MTSlive: SITUATION DETECTED: The White House Office of Science and Technology Policy has released a report directing federal research funding toward individual scientists and AI use rather than universities, per WSJ.

The directives will shape about $200B in annual federal R&D spending.

See 5 related tweets

@bubbleboi: I wonder where these independent scientific researchers work\n\nQT @AndrewCurran_: In an new report ...
@CSProfKGD: RT @Polymarket: BREAKING: White House to reportedly redirect "billions" in federal research funding ...
@Austen: I kind of love this\n\nQT @MTSlive: SITUATION DETECTED: The White House Office of Science and Techno...
@Techmeme: Memo: the White House OSTP plans to redirect federal research funding from universities to individua...
@JordanSchachtel: RT @AndrewCurran_: In an new report expected to be released later today the Trump administration wil...