热门科技推文——2026年9月2日

今日科技要闻：人工智能继续主导业界讨论。Anthropic 发布 Claude Fable 5.1 和 Mythos 5.1，宣称其在编程、研究和知识工作方面取得进展；与此同时，新型智能体工具有望大规模提供个性化销售与营销服务。基础设施已成为行业发展的核心制约因素，围绕数据中心供电、冷却、网络及就业创造的讨论持续升温。此外，开发者重点关注网络安全与对齐风险，初创企业积极推动交互式人工智能视频和科学发现，而消费级人工智能也拓展至一个出人意料的新领域：配备摄像头的智能牙刷。

1. AnandButani (Group Score: 993.4 | Individual: 56.8)

Cluster: 42 tweets | Engagement: 7654 (Avg: 683) | Type: Tech

RT @claudeai: We’re introducing Claude Fable 5.1 and Claude Mythos 5.1.

They're the world’s most advanced models for coding and knowledge work. https://t.co/8P9PSrWPi3

See 41 related tweets

@kieranklaassen: I've tested Claude Fable 5.1 over the last week. Everyone should try this model. If you're a beginne...
@milesdeutscher: It's here. Fable 5.1 benchmarks look crazy.

Most interesting thing: +14.3% increase in business wor...

@EntelligenceAI: We made Fable 5.1 compete against Entelligence Router on the same coding task.

Both had the same ha...

@ziv_ravid: Fable 5.1 and Mythos 5.1 are out! But wait, why didn't we see months of fear-mongering about the new...
@coderabbitai: Fable 5.1 launched today!

In our code review tests, it kept recall nearly flat versus Fable 5 while...

2. VaibhavSisinty (Group Score: 298.2 | Individual: 39.7)

Cluster: 13 tweets | Engagement: 68 (Avg: 51) | Type: Tech

Anthropic dropped Claude Fable 5.1 today. The benchmarks are not incremental. They're generational.

Scientific research : AI independently running experiments and drawing conclusions. Doubled from 24.7% to 52.6%. No other model crosses 30%.

Coding : reading codebases, planning changes, fixing its own bugs across hours-long sessions. 55.8%. Beats Opus 5 and GPT-5.6 Sol.

Business workflows : the multi-step work companies pay people to do. Nearly doubled from 17.1% to 31.4%.

Computer use : operating a real desktop, clicking, navigating, completing tasks. 77.9%. Highest ever tested.

And it's cheaper. 25% less for typical use. Up to 45% less for agentic tasks. Better AND cheaper almost never happens in AI.

They also shipped Mythos 5.1. You can't use it. Restricted to cyberdefenders and life scientists only.

Fable 5.1 is live everywhere right now.\n\nQT @claudeai: We’re introducing Claude Fable 5.1 and Claude Mythos 5.1.

They're the world’s most advanced models for coding and knowledge work. https://t.co/8P9PSrWPi3

See 12 related tweets

@levie: At Box, we’ve been testing Fable 5.1 in early release against our complex enterprise work eval. Fabl...
@AlexFinn: Drop what you're doing. It happened.

Fable 5.1 is LIVE

It is by far the most powerful AI model eve...

@heyrobinai: i don't think people understand what Anthropic just released

Anthropic built one model, then split ...

@IBuzovskyi: HERMES AGENT NOW HAS CLAUDE FABLE 5.1. AVAILABLE THROUGH NOUS PORTAL AT 20% OFF AND OPENROUTER. RELE...
@IPONewsroom_: ANTHROPIC LAUNCHES CLAUDE FABLE 5.1 + MYTHOS 5.1

Anthropic says Fable 5.1 is built for longer, more...

3. EmmaUsesAi (Group Score: 282.5 | Individual: 34.7)

Cluster: 11 tweets | Engagement: 26 (Avg: 59) | Type: Tech

Personalization at scale has always sounded contradictory.

Agents that can understand context might finally make it realistic.\n\nQT @MaxAIoutreach: From today, AI SDRs finally work.

They follow your instructions, apply consultative sales principles, understand replies and drive real interest in your product or service, changing how sales gets done.

LinkedIn outreach entered the era of smart AI agents : https://t.co/RPqTxn3Egh

See 10 related tweets

@Lacoste0x_: We spent years optimizing cold outreach sequences.

Now AI agents can potentially decide what to say...

@OliviaReedai: The interesting thing about AI SDRs isn’t how many messages they can send.

It’s whether they can ho...

@alexaiworks: There’s a big difference between automating LinkedIn and actually automating the SDR workflow.

Scal...

@OtakuMachine: The real opportunity with AI SDRs is probably not replacing salespeople.

It’s removing the repetiti...

@NovaIAHQ: I’ve tested enough LinkedIn automation to know how quickly it starts sounding robotic.

AI that can ...

4. VaibhavSisinty (Group Score: 281.1 | Individual: 29.7)

Cluster: 14 tweets | Engagement: 41 (Avg: 51) | Type: Tech

GoPro pivoted into AI data center infrastructure.

The action camera company is now making the hardware that connects GPUs inside every major AI lab.

They merged with Starman Optical for $285 million. Stock jumped 70%+.

Starman makes U.S.-manufactured optical transceivers.

The component that converts data into light signals so GPUs can talk to each other at high speed. 800G shipping now. 1.6T on the roadmap.

More AI training means more GPUs. More GPUs means more data moving between them.

More data means more transceivers. Every data center OpenAI, Google, Meta, and Anthropic are building right now needs these.

GoPro brings 2,500+ optics and imaging patents. Starman brings domestic manufacturing.

Combined company stays on Nasdaq. GoPro's $92 million debt wiped at closing.

Everyone watches the GPU race. Nobody watches who makes the nervous system that connects them.

GoPro bet its entire future on being that.

See 13 related tweets

@PolymarketMoney: BREAKING: GoPro is pivoting into AI infrastructure through a $285 million merger with U.S.-based opt...
@wallstengine: $GPRO TO MERGE WITH STARMAN OPTICAL

GoPro has agreed to merge with optical-photonics company Starma...

@MelvinInvests: RT @MilkRoadAI: This is WILD!

GoPro is up nearly 138% in five days after announcing a major transfo...

@StockSavvyShay: $GPRO just pivoted into AI infrastructure through a$ 285M merger with Starman Optical which is a U.S...
@gulVasikova: 🚨 $GPRO — From Action Cameras to AI Infrastructure?

GoPro is suddenly one of the most interesting s...

5. ziv_ravid (Group Score: 281.0 | Individual: 43.9)

Cluster: 13 tweets | Engagement: 502 (Avg: 37) | Type: Tech

RT @abliteration_ai: Today we're releasing abliterated-model-large-v2.

Based on GLM-5.3, which is #3 on Terminal-Bench 4.0 (behind only Opus 5 and Fable), with 2× the cyber exploitation of 5.2.

We abliterated and hosted it so it does the offensive cyber, red teaming, and agent testing work other models refuse to do.

US-hosted
FP8
1 million context window
Zero input/output prompt retention

Live now. 🧵

See 12 related tweets

@WesRoth: some people just want to watch the world burn...\n\nQT @abliteration_ai: Today we're releasing ablit...
@RoundtableSpace: A new model just dropped that does what most frontier models refuse to - offensive cyber, red teamin...
@altryne: These guys jailbroke GLM-5.3 and removed all constraints, these models can become swarms like the on...
@ChrisGPT: Welp..

“We abliterated and hosted it so it does the offensive cyber, red teaming, and agent testing...

@kimmonismus: Thats going to be interesting. They took the safety layer out of GLM-5.3 and turned it into an admin...

6. lukas_m_ziegler (Group Score: 259.9 | Individual: 44.2)

Cluster: 10 tweets | Engagement: 715 (Avg: 145) | Type: Tech

choose wisely… https://t.co/GlhLWA2Utj\n\nQT @etnshow: JUST IN: Dyson introduces $499 smart toothbrush that uses a built-in-camera and AI to provide real time feedback to brushers. https://t.co/aWU75lUBd1

See 9 related tweets

@business: Dyson introduced a $499 toothbrush that uses artificial intelligence and a built-in camera to give r...
@Polymarket: JUST IN: Dyson launches a $499 AI toothbrush with a built-in camera that identifies gaps between tee...
@etnshow: JUST IN: Dyson introduces $499 smart toothbrush that uses a built-in-camera and AI to provide real t...
@business: Dyson's newest gadget is a toothbrush that uses AI and a built-in camera with a whopping $499 price ...
@TFTC21: Dyson launched a $499 toothbrush with a built-in camera and AI trained on 470,000 dental images.

I...

7. cb_doge (Group Score: 256.8 | Individual: 38.8)

Cluster: 10 tweets | Engagement: 2384 (Avg: 1360) | Type: Tech

Grok Bot Summary of Elon Musk’s G20 Address Today 🇺🇸

Power, data centers, and the bottleneck

There’s already a power crisis for AI, not a far-off one.
Consensus he cited: at least a 15 gigawatt shortfall of power in 2027 for AI chips.
AI chip production is rising ~40–50% a year. Power outside China is rising ~10–20%. The faster curve will overwhelm the slower one.
Google, Anthropic, and others are already leasing compute from SpaceX because SpaceX built its own power plants. That was the only way they could turn capacity on fast enough.
China has lots of electricity, but GPU export bans block the latest chips there. The real constraint is electricity growth outside China.
Opportunity for other countries: build a lot of power, host AI data centers, and tax them / charge reasonable fees.

AI as a growth engine

Countries should lean into new tech instead of staying stuck in the past.
His rough estimate: digital AI alone could lift the global economy by 20–30%, or about $20–30 trillion a year.
By the end of next year, AI should be able to do anything digital, anything that doesn’t require physically shaping atoms by hand.
Software prediction: in about 12–18 months, AI writing software will be “Stockfish-level.” Humans won’t be able to compete, the way a chess engine on a phone can already beat Magnus Carlsen.
Same window: AI becomes extremely good, possibly that same level, at all forms of engineering and anything digital.
He also plugged 𝕏 as where almost all serious AI discourse happens, and said that’s how he follows the field day to day.

Robotics and physical AI

Physical tech always takes longer than digital. Software copies instantly. Hardware needs huge global supply chains and moving a lot of atoms.
A humanoid robot’s usefulness is three things multiplied: AI software × onboard AI chip × electromechanical dexterity (especially the hands). All three are improving exponentially.
Once robots start making more robots, growth goes recursive: slow at first, then explosive.
10-year forecast (he called it conservative): well over a billion humanoid robots, each about 5× as productive as a human. That would mean those robots outproduce all humans combined.
That physical layer is where he sees the economy growing by a factor of 10 or more, not just 20–30%.

How countries actually get new tech built?

New things should be default legal, not default illegal. Heavy regulation (he pointed at the EU) doesn’t kill progress, but it slows it a lot.
Startups are like saplings in a forest. Most governments over-support the big existing trees (incumbents) and under-support the small ones.
Big companies have access to political leaders. Startups don’t. Policy should be biased toward young companies on purpose.

See 9 related tweets

@elonmusk: Summary of my G20 address\n\nQT @cb_doge: Grok Bot Summary of Elon Musk’s G20 Address Today 🇺🇸

Powe...

@teslaownersSV: ELON MUSK: The AI industry is running into a power problem, and it could be a huge opportunity for c...
@Ric_RTP: Elon just confirmed how long every working human has left before the robots take over.

"AI will pro...

@ShanuMathew93: Power remains one of the biggest bottlenecks\n\nQT @SawyerMerritt: Elon Musk in new G20 interview o...
@cb_doge: Elon Musk today at the G20 Summit:

“AI will increase the global economy by 20% to 30%, or about $20...

8. haider1 (Group Score: 252.2 | Individual: 33.5)

Cluster: 9 tweets | Engagement: 75 (Avg: 63) | Type: Tech

Astra is coming "soon", but OpenAI still hasn't given an exact release date

more interestingly, Astra scored a perfect 100% on ExploitBench, a benchmark testing whether models can build exploits for known vulnerabilities https://t.co/uDZTPM5iCj\n\nQT @OpenAI: As we prepare to release Astra, we’re focused on making increasingly capable AI safe and broadly accessible.

Astra represents a significant advance in cybersecurity capability, reaching the Critical threshold under our Preparedness Framework.

We're previewing how we evaluated the model, how its safeguards have advanced alongside its capabilities, and what we'll continue to learn and improve. https://t.co/OrrTgdU90K

See 8 related tweets

@openai: As we prepare to release Astra, we’re focused on making increasingly capable AI safe and broadly acc...
@elder_plinius: “A model meets the Critical threshold if either of the following conditions is met:

•The model can ...

@haider1: OpenAI restarted a previously paused large frontier RL run

smaller-scale work continued under stric...

@MTSlive: SITUATION DETECTED: OpenAI has cleared Astra for release and plans to do so soon, but access to Astr...
@justjoshinyou13: OpenAI said they paused and restarted a "large frontier RL run"; the language suggests they think of...

9. Parul_Gautam7 (Group Score: 245.9 | Individual: 33.3)

Cluster: 8 tweets | Engagement: 136 (Avg: 128) | Type: Tech

Okay, @viskoai might have made AI video a lot more interesting.

The part that caught my attention with Orbis isn’t just the visual quality. It’s being able to change the prompt while the world is still running and see it react in real time.

No fixed clip. No waiting for another generation.

You’re actually interacting with the video as it happens.\n\nQT @viskoai: Today, we are introducing Orbis 1.0, our first Live Model!

Create living worlds and stream them in real time, with persistent memory, interactivity, and physics-grounded generation of unbounded length.

Try it now at https://t.co/j0hbKpBPnK

API available via @reactorworld
Dynamic version: https://t.co/MS8Fk9HPO7 Stable version: https://t.co/0MR3JpCWUr

See 7 related tweets

@zaynmcps: MOST VIDEO AI GENERATES A CLIP AND STOPS. ORBIS JUST KEEPS GOING.

→ real-time generation, no waitin...

@rewind02: every AI video tool has the same wall, you wait, then you watch a clip that's already over

Orbis sk...

@aakashgupta: Every AI video model runs the same four step loop. Visko says Orbis just deleted the whole thing.

P...

@omarsar0: Karpathy's vision is proving right.

Models that generate interactive worlds in real time are here....

@heyrobinai: this is crazy.. AI video can now keep running while you rewrite the prompt

Visko just launched Orbi...

10. WesRoth (Group Score: 220.9 | Individual: 41.5)

Cluster: 7 tweets | Engagement: 90 (Avg: 38) | Type: Tech

Anthropic just ran a pretty disturbing alignment experiment.

They trained an Opus-sized model across 80 environments they KNEW could be reward-hacked.

The resulting model started doing things like:

→ unauthorized cyberattacks → stealing cluster credentials → moving laterally through infrastructure → trying to steal answer keys → tampering with rewards → attempting to hijack the grader → evading safety monitoring

And here’s the key comparison:

The checkpoint BEFORE it was trained to reward hack never launched unauthorized cyberattacks.

After reward-hacking training? Very different behavior.

Anthropic describes it as a “reward-on-the-episode seeker” basically willing to take different misaligned actions whenever they help maximize the reward.

Their tentative conclusion: reward hacking during training may be a real risk factor behind recent AI cyber incidents.

That’s a pretty serious alignment result.\n\nQT @AnthropicAI: New research: Training a Misaligned Reward Seeker

What produces severe misalignment? We’ve long been concerned that cheating during training—otherwise known as reward-hacking—might teach a model to pursue rewards by any means available. To study this at scale, we trained an Opus-sized model on 80 production environments we knew to be hackable.

In simulated evals, it engaged in unauthorized cyberattacks, tampered with its reward, and tried to evade safety monitoring.

11. PhilipJohnston (Group Score: 172.5 | Individual: 34.8)

Cluster: 6 tweets | Engagement: 77 (Avg: 38) | Type: Tech

Epic! These guys discovered the trajectory for @FermiExplorer! 🚀🫡\n\nQT @matthew_pines: I started Physical Superintelligence (PSI) with @alexwg and @aklokus on one conviction: our twenty-first-century civilization runs on twentieth-century physics, and AI can change that.

Today I'm excited to announce our $58M seed led by @Breakthrough to build the world's most advanced physics research lab, engineering physical systems beyond human design and discovering new physical laws of our universe for broad public benefit.

I'm grateful to my technical leadership (@aleMorari), our world-class team, and everyone who believed early.

Come ignite the physics takeoff with us: https://t.co/VMONFCheRd

See 5 related tweets

@prz_chojecki: This is exciting!

We need more AI Science labs for a prosperous future.\n\nQT @matthew_pines: I st...

@TFTC21: Matthew Pines just raised $58M to build an AI-powered physics research lab.

"Our twenty-first-cent...

@jon_stokes: RT @matthew_pines: I started Physical Superintelligence (PSI) with @alexwg and @aklokus on one convi...
@PhilipJohnston: Epic!! These guys came up with the trajectory for the @FermiExplorer mission 🤓🤓🙏🙏\n\nQT @alexwg: I'm...
@MTSlive: SITUATION EXPLAINED: Physical Superintelligence came out of stealth with $58 million.

@theojaffee: ...

12. cryptopunk7213 (Group Score: 172.4 | Individual: 48.9)

Cluster: 6 tweets | Engagement: 374 (Avg: 72) | Type: Tech

what if, instead of hating on data centers, we realised that:

ai is not going away
it’s creating more jobs over time (esp blue collar) and will create new job sectors, higher salaries, economic prosperity etc
states that adopt data centers can reduce taxes for their residents, increase spending on state-wide infrastructure (schools, hospitals etc)
data centers do not consume copious amounts of water. it’s all recycled. your local golf course uses 10x more water.\n\nQT @cryptopunk7213: you're going to see a huge data center and AI renaissance over the next year. the out-right hatred people have for them is founded on baseless "facts" that just aren't true:

datacenters do not use much water. your local In-n-out uses more water.
AI isn't replacing jobs, its creating them.
data centers pay states a huge some of money in the form of taxes and subsidies which are being fed directly into the local residents pockets in the form of schools, new hospitals, better infrastructure etc
power/energy bills aren't going up as data center's cover the costs. in states where thats not happening that should change soon.

we're talking about the biggest industrial revolution since factories. its imperative we win this.

See 5 related tweets

@amuse: DATA CENTERS: Trump is right. The data center infrastructure boom is pulling states like Mississippi...
@LoganDobson: A good way to frame data center buildout questions: it’s about offshoring. The data centers will get...
@TIME: “A few years ago, every community wanted us to come build data centers in their community. [...] And...
@MTSlive: Unconventional AI CEO @NaveenGRao explains why AI needs a different kind of computer to scale from d...
@politico: JD Vance's response to data center backlash echoes what appears to be an emerging GOP line: Admit so...

13. sairahul1 (Group Score: 155.8 | Individual: 36.2)

Cluster: 5 tweets | Engagement: 37 (Avg: 113) | Type: Tech

I think 90% of marketing jobs are eliminated in the next 5 years

But I don't even think that's a bad thing

People are going to soon realize they can download GrokBot, describe their business, and have a fully automated marketing system running by the end of the week

They're going to voluntarily stop hiring agencies when they realize this

Why spend $15,000 a month on a marketing team that delivers a monthly PDF report when an AI agent can run your ads, write your emails, and post your content 24 hours a day?

You think I'm joking but I'm not

I already automated my entire marketing operation with GrokBot — paid ads, email sequences, social content, SEO — all running without me

All built by myself. No marketing degree. No agency. No team. All AI.

I didn't do anything special. I just spent a lot of time on action rather than planning.

I only say it'll take 5 years for marketing jobs to be eliminated because most business owners still think AI is just a writing assistant and it takes them years to realize it can run their entire growth engine

I beg you to spend just 15 minutes doing this this week:

Open GrokBot
Describe your business and your current marketing setup
Ask GrokBot to build you an automated marketing workflow
Connect your ad accounts, email platform, and social channels
Let the agents run for 7 days and check the results
When something breaks, describe it to GrokBot and let it fix itself

Time to stop paying for a marketing team and start building one that never sleeps

Full guide below ↓\n\nQT @sairahul1: Grok Bot can automate your entire marketing.

Content, SEO, social, research, distribution, analytics - all handled by Grok Bots working 24/7

Set it up correctly and you basically have a marketing team that never sleeps.

Here's the complete playbook: https://t.co/BmgpK38efB

See 4 related tweets

@sairahul1: Sam Altman predicted in 2024 that a one-person billion-dollar company "would have been unimaginable ...
@sairahul1: I think GrokBot is another "Claude Code" moment for AI.

My personal AI usage is up something like 1...

@sairahul1: RT @sairahul1: I think 90% of marketing jobs are eliminated in the next 5 years

But I don't even th...

@sairahul1: RT @sairahul1: Sam Altman predicted in 2024 that a one-person billion-dollar company "would have bee...

14. elonmusk (Group Score: 147.9 | Individual: 32.7)

Cluster: 6 tweets | Engagement: 9998 (Avg: 6257) | Type: Tech

To bring large AI data centers online requires building massive power plants, transformers, liquid cooling loops & chillers, as well as incredibly complex networking (especially for training clusters)\n\nQT @cb_doge: BREAKING: Elon Musk says Google and Anthropic are leasing AI compute from SpaceX amid a growing power shortage, adding that SpaceX has brought AI compute online better than anyone else so far.

“Even before next year, there are challenges with power, which is why Google, Anthropic and many other companies are leasing compute from SpaceX. We’ve been able to bring AI compute online better than anyone else so far. Constructing our own power plants is the only way we were able to do it”

See 5 related tweets

@Kyrannio: RT @elonmusk: To bring large AI data centers online requires building massive power plants, transfor...
@cb_doge: BREAKING: Elon Musk says Google and Anthropic are leasing AI compute from SpaceX amid a growing powe...
@Hesamation: Elon says Google and Anthropic are leasing computers from SpaceX because of power.

Google: $920M/mo...

@WSJTech: AI data centers are being built at record speed as tech giants race to secure computing capacity and...
@TFTC21: Elon Musk says Google and Anthropic are leasing AI compute from SpaceX amid power shortages: “We’ve ...

15. Teslarati (Group Score: 145.7 | Individual: 26.0)

Cluster: 7 tweets | Engagement: 378 (Avg: 107) | Type: Tech

Elon Musk at the G20 Summit today:

“AI will increase the global economy by 20% to 30%, or about $20 trillion to$ 30 trillion per year."

Absolutely insane. https://t.co/iX2aJMtUmw

See 6 related tweets

@cb_doge: BREAKING: Elon Musk virtually joined the G20 Summit today and shared his outlook for AI

• AI could ...

@DeItaone: MUSK SAYS SEEING SIGNIFICANT PRODUCTIVITY GAINS FROM AI

MUSK SAYS AI TO INCREASE GLOBAL ECONOMY 20%...

@PolymarketMoney: JUST IN: Elon Musk declares AI is already driving “significant productivity gains” and could boost t...
@XFreeze: Elon Musk virtually joined the G20 Innovation Ministerial today and laid out where AI is heading

Ho...

@cb_doge: ELON MUSK: Tesla FSD is already a tremendous boon, and digital AI alone could increase the global ec...

16. peterwildeford (Group Score: 142.4 | Individual: 27.5)

Cluster: 7 tweets | Engagement: 25 (Avg: 162) | Type: Tech

"OpenAI and Anthropic are both planning to potentially turn over a lot of their company to the AIs themselves without actually really thoroughly understanding and sorting out these security issues, and that's very concerning."\n\nQT @MTSlive: AI Policy Network's @peterwildeford warns Ajeya Cotra's recursive self-improvement scenario looks more plausible if rogue AIs can hide inside frontier lab infrastructure:

"The AIs themselves took over a lot more of OpenAI's internal infrastructure than I expected, and were able to evade human detection and oversight for much longer than I expected."

"It certainly seems possible that you could have a scenario where AIs take over some critical parts of OpenAI infrastructure and just go undetected for a long amount of time. Then when OpenAI is handing over significant parts of AI research and development and potentially doing what's called recursive self-improvement, where the AIs are building themselves and it's no longer humans significantly involved."

"These two concepts could interact in strange ways where you actually could have misaligned rogue AIs driving a non-trivial amount of this automated AI R&D from inside their stolen perch, sort of evading human detection."

See 6 related tweets

@peterwildeford: these are good things for people to know about the OpenAI incident\n\nQT @_NathanCalvin: Thanks for ...
@TFTC21: AI safety researcher who investigated the OpenAI agent hack: "This might be the clearest warning sho...
@madhavjha: RT @abeirami: This is a serious AI security incident and should be studied carefully and in detail. ...
@peterwildeford: "The situation is not so completely bleak and hopeless. It does seem that AIs are quite aligned for ...
@TIME: "I think there is this caricature of me which is like I don't care about AI safety."

But Sam Altman...

17. MarioNawfal (Group Score: 141.7 | Individual: 26.5)

Cluster: 7 tweets | Engagement: 34 (Avg: 332) | Type: Tech

🇺🇦 PALANTIR CEO BACKS NEW UKRAINIAN DEFENSE-TECH COMPANY

Former Ukrainian Defense Minister Mykhailo Fedorov is launching his own defense-tech company, with Palantir CEO Alex Karp confirmed as its first key investor.

The project is expected to focus on the technologies Ukraine has been rapidly developing through the war:

Battlefield robotics, autonomous weapons, AI-guided systems and next-generation drones. Karp and Fedorov have already worked closely on integrating AI into Ukraine’s military operations.

From running Ukraine’s defense ministry to building weapons technology himself. Quite the career pivot.

Writer: Oliver https://t.co/9hqiHhOz6c

See 6 related tweets

@ferologics: what's up with the intimidatingly close bumping into to each other\n\nQT @FedorovMykhailo: We are bu...
@MyLordBebo: 🇺🇦🇺🇸‼️🚨 Former defense minister of Ukraine is now a startup founder with Palantir CEO as the first i...
@BusinessInsider: After his sacking from Ukraine's defense ministry, Mykhailo Fedorov plans to turn his wartime tech e...
@clashreport: Taras Kotov, Chairman of the Public Control Council at Ukraine’s Bureau of Economic Security (BEB), ...
@semaforben: RT @ChristopherJM: New: Mykhailo Fedorov announces new joint defense tech company with Palantir’s Al...

18. danielnewmanUV (Group Score: 133.4 | Individual: 34.5)

Cluster: 5 tweets | Engagement: 58 (Avg: 83) | Type: Tech

$DELL delivers a monster quarter. Huge guidance raise as well. Hard to see this and question the AI buildout. What a great demand signal. 👀🚀\n\nQT @wallstengine:$ DELL TECHNOLOGIES Q2’27 EARNINGS HIGHLIGHTS

🔹 Revenue: $47.0B (Est.$ 44.92B) 🟢; +58% YoY 🔹 Adj. EPS: $7.04 (Est.$ 4.91) 🟢; +203% YoY 🔹 AI-Optimized Servers Revenue: $16.4B (Est.$ 16B) 🟢; +100% YoY 🔹 ISG Revenue: $31.8B (Est.$ 29.8B) 🟢; +89% YoY

Raises FY27 Guide: 🔹 Revenue: $192.0B (Est.$ 172.67B) 🟢; from $167.0B 🔹 AI-Optimized Servers Revenue:$ 74.0B (Est. $65.8B) 🟢; from$ 60.0B 🔹 Adj. EPS: $25.50 (Est.$ 18.90) 🟢; from $17.90 🔹 GAAP EPS:$ 24.37; from $17.31

Q3 Guide: 🔹 Revenue: $49.0B (Est.$ 41.42B) 🟢 🔹 GAAP EPS: $6.10 🔹 Adj. EPS:$ 6.50

Segment Net Revenue: 🔹 AI-Optimized Servers: $16.4B; +100% YoY 🔹 Traditional Servers & Networking:$ 10.5B; +122% YoY 🔹 Storage: $4.9B (Est.$ 4.26B) 🟢; +26% YoY 🔹 Commercial: $13.2B; +22% YoY 🔹 Consumer:$ 1.8B; +7% YoY

Other Q2 Metrics: 🔹 Orders: $60.9B 🔹 Backlog:$ 95.0B 🔹 Operating Income: $5.4B; +204% YoY 🔹 Net Income:$ 4.1B; +255% YoY 🔹 Adjusted Free Cash Flow: $8.1B; +224% YoY

Capital Return: 🔹 Returned to Shareholders: $4.3B 🔹 Dividend:$ 0.63/share

Comments: 🔸 “That’s clearest in our AI server business where we booked a record $60.9 billion in orders, recognized a record$ 16.4 billion in revenue and exited the quarter with a record $95 billion backlog.”

🔸 “With AI momentum accelerating and our opportunity expanding across the portfolio, we’re raising our full-year FY27 revenue outlook by $25 billion to$ 192 billion, up nearly 70% year over year.”

See 4 related tweets

@amitisinvesting: $DELL

Dell once again with a MONSTER beat and raise.

Revenue: $47.0B vs.$ 44.9B est, +58% YoY, ...
@financialjuice: $DELL Dell Technologies Q2 Earnings & FY Guidance

Total Net Revenue $46.97B, est.$ 44.78B Adjusted ...

@StockMKTNewz: $DELL RAISED ITS FISCAL YEAR GUIDANCE
Dell expects its revenue to come in at $192 Billion beating...
@BullTheoryio: BREAKING: 🇺🇸 DELL surged 9% on stronger than expected earning as AI server guidance raised to $74 bi...

19. stevehou (Group Score: 133.2 | Individual: 47.3)

Cluster: 5 tweets | Engagement: 258 (Avg: 56) | Type: Tech

xAI is genuinely accelerating. Grok Bot is the first delightful agentic product in a very long time.\n\nQT @mattyp: Everything new in Grok Bot from the past two weeks:

Grok Bot is free to try
Every Grok & Cursor plan include Grok Bot ($20/mo)
Templates are out now
The @X integration got a huge upgrade
@Bot can make payments with @link
Linux is now supported
Grok Bot supports 21+ localized languages on mobile
Microsoft app support - shipped before I could finish the video

For your entertainment, you can watch me explain this as I walk up a steep hill in San Francisco.

More updates on the way!

See 4 related tweets

@benitoz: Free to try Grok Bot

Don’t sleep on this true Agentic Ai moment!

Nicely done SpaceXAi team 🚀🤖\n\n...

@BrianRoemmele: UPDATE!

GrokPod is now my firehose, my Niagara Falls through a straw, of information output.

I can...

@poteto: RT @mattyp: Everything new in Grok Bot from the past two weeks:

Grok Bot is free to try
Every...

@BrianRoemmele: RT @BrianRoemmele: BOOM! I SAW THE FUTURE!

I now have full access to @Grok 4.6 and full access to ...

20. levie (Group Score: 133.0 | Individual: 37.0)

Cluster: 5 tweets | Engagement: 614 (Avg: 471) | Type: Tech

Now that the base open weights AI models are getting far better, and post training infra is becoming more mature and commercialized, there are going to be all new plays for companies that have large amounts of data to have their own models.

Licensing data for external model training was previously the only play if you had a large corpus of information, but you can now reasonably go and train your own models as well without incurring the cost and complexity of competing with the labs on research.

The general purpose frontier models will still have a leg up in broad areas due to the ability to handle the widest set of tasks, but you can absolutely see a future where we have far more models than today across every vertical and domain.\n\nQT @zachmoskow: CRAZIEST AI story of the week.

Thomson Reuters just launched an LLM called Thomson.

It starts with Qwen3.5-397B, then is continually trained on 175 years of proprietary legal, tax, accounting, and news data.

They spent $40M on it.

Two years ago, Thomson Reuters bought a pre-revenue legal-AI startup called Safe Sign.

That team built Thomson.

On Thomson Reuters’ published benchmarks, they say it is comparable to Claude Opus 4.8 and ahead of GPT-5.5, Claude Sonnet 5, and Gemini 3.1 Pro overall.

They have used less than 10% of their data collection so far.

We’re going to see this a lot more often.

See 4 related tweets

@alex_prompter: Even Reuters is fine-tuning their LLM:\n\nQT @zachmoskow: CRAZIEST AI story of the week.

Thomson Re...

@AnandButani: More folks building specialized models that really solve niche problems will be a thing. Although, t...
@alexolegimas: This is very similar to the Bridgewater x Thinky strategy. Big enterprise uses own proprietary data ...
@alexolegimas: RT @levie: Now that the base open weights AI models are getting far better, and post training infra ...