科技热点推文 - 2026年6月13日

科技每日简报 | 2026年6月13日

Today's top tech conversations are led by @MiniMax_AI, whose post about 'MiniMax M3, Open-Weight, Now O...' garnered the highest engagement. Key themes trending across the top stories include spacex, today, https, years, bloom. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. MiniMax_AI (Group Score: 450.5 | Individual: 64.8)

Cluster: 15 tweets | Engagement: 2825 (Avg: 157) | Type: Tech

MiniMax M3, Open-Weight, Now On Hugging Face , with only ~428B parameters and ~23B activated parameters

Weights: https://t.co/g4Ybfa2kWH MiniMax Sparse Attention: https://t.co/HcTlWRotG3\n\nQT @MiniMax_AI: Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capabilities

Coding & Agentic Frontier: 59.0% SWE-Bench Pro, 66.0% Terminal Bench 2.1, 34.8% SWE-fficiency, 28.8% KernelBench Hard, 74.2% MCP Atlas
MiniMax Sparse Attention scales context to 1M
Natively Multimodal from Step Zero

API: https://t.co/fHRdSV7BwZ Token Plan: https://t.co/BDCycxepZw 🚀New! MiniMax Code: https://t.co/GvB4YiB6Ul

Weights & Tech Report in ~10 Days

See 14 related tweets

@yacineMTB: RT @MiniMax_AI: Introducing MiniMax M3: The First Open-Weights Model to Combine Three Frontier Capab...
@NVIDIAAI: Congrats to the @MiniMax_AI team on the release of MiniMax M3, a long-context multimodal model for t...
@FireworksAI_HQ: MiniMax M3 is live on Fireworks. Day-0, fastest endpoint for the MiniMax series.

→ Top open-weight ...

@yacineMTB: https://t.co/2D8qJfJrvj\n\nQT @MiniMax_AI: Introducing MiniMax M3: The First Open-Weights Model to C...
@huggingface: RT @MiniMax_AI: MiniMax M3, Open-Weight, Now On Hugging Face , with only ~428B parameters and ~23B a...

2. teortaxesTex (Group Score: 438.1 | Individual: 52.7)

Cluster: 17 tweets | Engagement: 469 (Avg: 46) | Type: Tech

I want to see this compared with Composer 2.5 Like, really hard Cursor has a ton of proprietary data, a large head start, and threw a Colossus at RLing Kimi K2.5 checkpoint. What is the gap now?\n\nQT @Kimi_Moonshot: 🌘 Kimi-K2.7-Code, our latest coding model, is now released and open-sourced!

🔷 Improved coding & agent performance over K2.6: +21.8% on Kimi Code Bench v2, +11.0% on Program Bench, and +31.5% on MLS Bench Lite. 🔷 Reasoning efficiency: Less overthinking, with 30% lower reasoning-token usage compared to K2.6. 🔷 Long-horizon coding: Improved instruction following, higher end-to-end coding task success rates.

⚡️ 6x High-Speed Mode coming soon! 🔌 Available today via Kimi API and Kimi Code.

🔗 Kimi Code: https://t.co/uvoSJKyGCY 🔗 API: https://t.co/EOZkbOwCN4

See 16 related tweets

@QuixiAI: RT @Kimi_Moonshot: 🌘 Kimi-K2.7-Code, our latest coding model, is now released and open-sourced!

🔷 I...

@lmsysorg: 🎉 Meet Kimi-K2.7-Code from @Kimi_Moonshot, their latest open-source coding model, day-0 support is n...
@ZenMuxAI: Kimi K2.7 is a serious jump — better coding, less overthinking, 30% fewer reasoning tokens.

Already...

@cgtwts: Kimi just dropped K2.7-Code

open sourced 21.8% better on Kimi Code Bench v2 11.0% better on P...

@MatthewBerman: Nearly frontier open source coding model. Congrats to the Kimi team.

I would love to see this teste...

3. BullTheoryio (Group Score: 164.3 | Individual: 42.4)

Cluster: 5 tweets | Engagement: 3333 (Avg: 1207) | Type: Tech

A man who earned $28 an hour at SpaceX is about to become a millionaire today because of the IPO.

But he is just 1 of 4,400 SpaceX employees becoming millionaires today.

SpaceX begins trading on Nasdaq today under the ticker SPCX, priced at $135 per share and valued at$ 1.77 trillion.

The offering raised $75 billion and drew more than$ 250 billion in investor demand, more than three times oversubscribed.

More than 4,400 current and former SpaceX employees are projected to become millionaires today. Around 400 of them will hold stakes worth over $100 million each.

To put that in context, the Google IPO in 2004 created roughly 1,000 millionaires. Facebook's 2012 listing produced a similar number. SpaceX is doing more than four times both of them in a single day.

And the majority of these people are not founders or executives.

They are engineers, welders, machinists, and launch technicians who took equity over cash and held it for years.

Juan Hernandez moved from Mexico and learned to weld for the money.

He joined SpaceX in 2015 earning $28 an hour. When he went full time, the company gave him$ 10,000 worth of stock. He held it, bought more through payroll deductions, and sold a small portion in 2020 to buy property in Texas.

His remaining shares are worth $880,000 today.

Trevor Hise joined as a launch engineer straight out of college in 2011. He spent 12 years accumulating shares through salary, bonuses, and reinvestment. He is 37 years old and walking away with more than $13.5 million.

Gavin Petit started in 2012 on an $80,000 base salary with a small initial grant priced at$ 13.80 per share. He took every bonus in stock.

He has already paid off his home in Denver from earlier tender offer windows and now holds a stake worth several million dollars.

J. André Lavoie, an engineer who received grants years ago, is sitting on more than $28 million. He is currently in Italy renovating a hotel.

Tom Mueller was SpaceX's first employee, hired in 2002 to build the engines that made all of this possible. He left in 2020 but kept his equity.

This week he said: "Elon always said that your salary is one thing, but it's the equity that's gonna be worth something. That day is here."

This is what SpaceX's compensation model actually looked like from the beginning. The company paid below market salaries and made up the difference in equity at every level of the organisation, not just at the top.

Stock options vested over four to six years. Workers could buy additional shares at a 15% discount through a company purchase plan.

Early option grants were priced below $2 per share. Even 2025 grants carried exercise prices between$ 37 and $42. At$ 135 today, those positions are sitting at 3x returns minimum.

The physical impact is already visible in Brownsville, Texas, one of the poorest cities in the United States, where more than 3,000 SpaceX employees work at Starbase.

Median home prices in the surrounding county have gone from $131,000 in 2014 to over$ 281,000 today.

More than 100 SpaceX employees pooled their combined holdings, estimated between $1 billion and$ 5 billion to negotiate institutional level wealth management fees as individual workers.

Today, SPCX opens and 4,400 people find out what years of holding actually pays.

See 4 related tweets

@heyshrutimishra: Elon just created 4,400 millionaires in a single day.

400 of them are now worth over $100 million. ...

@MilkRoadAI: RT @MilkRoadAI: This is WILD!

The most important story coming out of the SpaceX IPO this morning is...

@shanaka86: Google turned $900 million into$ 134 billion.

That is the filing math at Friday's peak. The Alaska ...

@heyshrutimishra: BREAKING: SpaceX is officially public 🚨

The largest IPO in history just made 4,400 employees millio...

4. peakxvpartners (Group Score: 158.9 | Individual: 37.7)

Cluster: 5 tweets | Engagement: 69 (Avg: 37) | Type: Tech

Extreme affordability is the biggest unlock for AI adoption in India.

Today, @Avataar launched Varya, with support from the #IndiaAIMission: India’s first distilled video model, demonstrating that AI video can be made 10x more efficient.

Built for India’s scale: ⚡ 4 generation steps vs 50 for Wan 2.2 🎥 5-sec 720p video in 45 secs vs 1,230 secs 💸 ₹0.48/sec hosted price 🚀 10x faster, at a fraction of the cost

The core differentiation for Indian models will be affordability, localization and cultural relevance.

Huge congratulations @AluruSravanth and team! 👏

Read more in @TechCrunch 👇 https://t.co/gBc8U3e3a3

@officialIndiaAI @secretarymeity @goi_meity @RajanAnandan @brettsingh

See 4 related tweets

@Rajananandan: Population scale adoption of AI in India requires "extreme affordability", local languages and cultu...
@TechCrunch: Cheaper, faster, and culturally aware, Avataar’s video AI is built for India’s scale https://t.co/bt...
@moneycontrolcom: #Business | India won't be second to anybody in developing foundational AI models: IT Secretary

-In...

@techdaily24: Cheaper, faster, and culturally aware, Avataar’s video AI is built for India’s scale https://t.co/xQ...

5. dkundel (Group Score: 154.1 | Individual: 29.3)

Cluster: 6 tweets | Engagement: 65 (Avg: 198) | Type: Tech

Combining this with /goal is the ultimate move for measurable performance improvements for your web apps\n\nQT @OpenAIDevs: Introducing developer mode for browser use in Chrome and the Codex in-app browser.

Codex can use the Chrome DevTools Protocol (CDP) to debug browser issues by profiling JavaScript performance and inspecting console output, network traffic, and page state. https://t.co/JTFjgCHmgI

See 5 related tweets

@jxnlco: Great ship by the browser team! Make sure to give @JamesZmSun a follow!\n\nQT @OpenAIDevs: Introduci...
@PaulSolt: More debugging in Codex for web based apps.

New Browser Use developer mode unlocked!

Check it out....

@dkundel: RT @OpenAIDevs: Introducing developer mode for browser use in Chrome and the Codex in-app browser.

...

@JamesZmSun: Two browser updates for today's Thursday Codex drop: Dev Mode + faster browser use.

1/ Dev Mode Now...

@ajambrosino: more thursday!\n\nQT @JamesZmSun: Two browser updates for today's Thursday Codex drop: Dev Mode + fa...

6. chandrarsrikant (Group Score: 142.3 | Individual: 35.1)

Cluster: 5 tweets | Engagement: 80 (Avg: 791) | Type: Tech

Who is Vik Bajaj? Meet the co-founder of Prometheus, Jeff Bezos's new AI startup

Thirteen years ago, Indian-origin scientist and tech executive Vikram "Vik" Bajaj worked closely with Google co-founder Sergey Brin at Google X, often referred to as the company's "moonshot factory", where he helped launch the tech giant's life sciences unit. Now, the MIT alumnus Bajaj is working with another legendary tech titan: Amazon founder Jeff Bezos.

Together they are leading the industrial AI startup Prometheus, which has raised $12 billion at an eye-popping$ 41 billion valuation. The startup, which seeks to build an "artificial general engineer," launched in November last year with $6.2 billion in funding. Bezos and Bajaj started working on Prometheus in late 2024.

Nicely done by @tsuvik

https://t.co/EsQOvQYjNB

See 4 related tweets

@WesRoth: Jeff Bezos’ AI startup Prometheus raised $12B in new funding at a roughly$ 41B valuation.

The compa...

@chandrarsrikant: RT @IPONewsroom_: JEFF BEZOS JUST EMERGED FROM STEALTH WITH A $41 BILLION AI STARTUP CALLED PROMETHE...
@moneycontrolcom: Who is Vik Bajaj? Meet the co-founder of Prometheus, Jeff Bezos's new AI startup
Vik Bajaj co-lea...
@tsuvik: 🚨Vik Bajaj, who previously co-founded Google's life sciences unit, is leading the physical AI startu...

7. Cloudflare (Group Score: 136.5 | Individual: 57.7)

Cluster: 3 tweets | Engagement: 1892 (Avg: 228) | Type: Tech

RT @Mastercard: As AI agents begin to act, payments move into the background — at machine speed and massive scale.

Today we’re introducing Mastercard Agent Pay for Machines — bringing structure, governance, and trust to this new class of payments.

Launching with 30+ partners to bring this to life from day one.

This isn’t just more payments. It’s a new operating model for commerce.

👉 Learn more: https://t.co/TeS6Lj8jLO

See 2 related tweets

@Cloudflare: Cloudflare 🤝 Mastercard\n\nQT @Mastercard: As AI agents begin to act, payments move into the backgro...
@Cloudflare: The internet was built for human interactions, but the infrastructure of the future must be built fo...

8. StockSavvyShay (Group Score: 136.2 | Individual: 32.1)

Cluster: 7 tweets | Engagement: 889 (Avg: 747) | Type: Tech

$NVDA has told Chinese customers its next-generation Vera AI data center CPU could launch as early as August.

Nvidia expects Vera to become a major new business targeting $20B in revenue by the end of this fiscal year. https://t.co/AIpIJR09x2

See 6 related tweets

@zephyr_z9: I wonder what the adoption will be Could be a good NVLink Fusion play (Chinese ASIC/GPU + Vera CPU a...
@Reuters: Exclusive: Nvidia begins Vera CPU sales pitch to Chinese clients, sources say https://t.co/VILsPxHpN...
@Techmeme: Sources: Nvidia has told Chinese clients that its new Vera CPUs for AI data centers could be availab...
@Reuters: Exclusive: Nvidia has told Chinese clients that its new ‘Vera’ central processors for AI data center...
@Cointelegraph: 🇨🇳 UPDATE: NVIDIA is pitching its Vera AI CPU to Chinese clients as it expands beyond GPUs into agen...

9. DeItaone (Group Score: 132.2 | Individual: 41.2)

Cluster: 4 tweets | Engagement: 1834 (Avg: 747) | Type: Tech

$SPCX - SPACEX IPO: THE BIGGEST BET IN MARKET HISTORY

SpaceX is really three businesses under one stock:

Starlink is the profit engine, generating $11.4B in revenue and$ 4.4B in operating profit in 2025. Subscribers grew from 2.3M in 2023 to over 10M by early 2026. SpaceX Launches generated $4.1B in revenue but lost money due to massive investment in Starship, the rocket intended to dramatically lower launch costs. AI (xAI + X) generated$ 3.2B in revenue but lost $6.4B in 2025, consuming all of Starlink's profits and more.

Without AI, SpaceX was profitable. With AI, it lost nearly $5B in 2025 and another$ 4.3B in Q1 2026 alone.

The IPO values SpaceX at $1.77 trillion, but only 4.3% of shares will trade initially. That limited float could drive strong early demand and volatility while making true price discovery difficult.

Retail investors are getting access to just 1.3% of the company, while insiders and major funds remain largely locked up. Those restrictions begin easing within months, potentially increasing selling pressure.

SpaceX will not immediately join the S&P 500, as it currently fails key inclusion requirements, delaying an estimated $14B of passive index-fund buying.

The biggest risk is valuation. At roughly 90x sales, SpaceX is trading at a premium far above the largest technology companies. Historical IPO research shows that highly valued, low-float, unprofitable IPOs often deliver strong first-day gains but weaker long-term returns.

The bullish case is simple: SpaceX dominates satellite internet and commercial launches, and Starship could reshape the economics of space. The bearish case is that investors are paying an unprecedented price today for profits that may arrive years from now.

Bottom line: SpaceX may become one of the most important companies of the century. But this IPO looks designed to maximize demand in the short term, while much of the future selling pressure arrives later. For retail investors, the excitement is obvious—the risk is whether today's valuation already prices in most of the future success.

See 3 related tweets

@cryptorover: People are treating SpaceX like a once in a lifetime opportunity.

But if you invest $11,000 today, ...

@rohanpaul_ai: RT @Meer_AIIT: 🚨News: SpaceX just made history with the largest iPO ever, and Elon musk has become t...
@eng_khairallah1: RT @mikenevermiss: SpaceX is listing on Nasdaq under $SPCX at$ 135/share, targeting a $1.75 trillion...

10. MilkRoadAI (Group Score: 129.2 | Individual: 32.7)

Cluster: 5 tweets | Engagement: 15 (Avg: 69) | Type: Tech

Bloom Energy $BE is the most capex-sensitive name in our AI PRO analyst's book.

It's a direct bet on the AI buildout running at full speed.

If spending plans slip, it gets hit first and hardest.

So on June 9, he trimmed. He took profits on part of the position, held the rest, and kept the 800V DC thesis intact.

The next day, Bloom dropped 8.6%.\n\nQT @MilkRoadAI: Our PRO analyst @whitecollarexit called Bloom Energy $BE on Feb 17, at ~$ 131 a share.

The stock ran as high as $322.83, a 146% gain from his entry - but on Tuesday, he started selling.

Here's why (save this)...

His original thesis was pretty straight forward:

AI data centers need huge amounts of electricity, and the public grid takes 3 to 5 years to deliver it.

Bloom's fuel cells put power on site in about 90 days.

Hyperscalers were never going to wait - they were going to pay for speed.

And the market caught on to this fast.

Bloom posted 130% revenue growth in Q1, signed a supply deal with Oracle, and saw its stock jump 20% on the report.

By late April it was @whitecollarexit's top-performing holding, up more than 100% since he posted his entry to PRO members.

He held anyway, because he believed the trade had a second leg...

Bloom's platform runs on 800V DC, the same power architecture NVIDIA's Rubin Ultra rack systems are expected to need in 2027.

As AI racks grow from 0.1 MW toward 1 MW each, older power setups won't be able to keep up.

@whitecollarexit argued Bloom was built for where the hardware roadmap was headed, and that the market hadn't priced that part yet. So he sat through the double.

But over the past few weeks, the data started to shift...

@whitecollarexit flagged five factors driving the change:

GPU utilization is reportedly running near 5%.
More than 60% of planned 2027 data center capacity hasn't started construction.
The Silicon Data index shows the frontier model premium stalling.
DeepSeek cut token costs by 75%.
Ramp's spending data shows buyers moving to cheaper open-source models.

In his view, none of that kills the power thesis.

What it does is raise the odds that the market starts questioning how much data center capex actually shows up in the near term.

Bloom is the most capex-sensitive name in his book.

It's a direct bet on the AI buildout running at full speed.

If spending plans slip, it gets hit first and hardest.

So on June 9, he trimmed. He took profits on part of the position, held the rest, and kept the 800V DC thesis intact.

The next day, Bloom dropped 8.6%.

The timing came from a framework @whitecollarexit has been writing about for months:

→ The AI buildout is a sequence of bottlenecks. → The market already repriced chips and memory. → Power is getting repriced now.

Once one layer gets fully priced in, the money moves to the next one.

The cash from this trim is already earmarked for that next layer.

PRO members saw every step of this trade in real time:

The February entry, the exact position size, the reasoning for holding through +100%, and the trim before the recent drop.

They'll also be first to see where the freed-up capital goes.

If you want that alert the moment it hits, the link is in the first comment below 👇

See 4 related tweets

@MilkRoadAI: AI data centers need huge amounts of electricity, and the public grid takes 3 to 5 years to deliver ...
@MilkRoadAI: On June 9, our PRO analyst @whitecollarexit trimmed his best performing asset (Bloom Energy $BE).

H...

@MilkRoadAI: RT @MilkRoadAI: Over the past few weeks, the data behind Bloom Energy has started to shift...

Our P...

@MilkRoadAI: RT @MilkRoadAI: On June 9, our PRO analyst @whitecollarexit trimmed his best performing asset (Bloom...

11. Rasmic (Group Score: 129.2 | Individual: 51.2)

Cluster: 3 tweets | Engagement: 617 (Avg: 119) | Type: Tech

I'm ngl

I'm mad i didn't think about this first\n\nQT @andrewmccalip: Get paid to wait

The Claude Code spinner might be the most watched line on Earth.

So I turned it into an ad marketplace.

Advertisers bid on it. You keep 50% of the money.

Install the extension → get cash from ads.

Introducing Kickbacks https://t.co/hcWRQAqKL2

See 2 related tweets

@anshnanda: I tried this so you don’t have to.

At the end, I got:

10,000 impressions
$600 spent
0-400 cli...
@shiri_shh: $1500 for 1K impressions 😭 https://t.co/6yYF4zKzty\n\nQT @andrewmccalip: Get paid to wait

The Claud...

12. jerryjliu0 (Group Score: 129.0 | Individual: 34.7)

Cluster: 4 tweets | Engagement: 50 (Avg: 66) | Type: Tech

There's got to be some major advances in the data ingestion, indexing, and retrieval layer to make this happen.

Agents are getting much more sophisticated in fetching diverse sources of context with simple tool interfaces (e.g. CLI + filesystem). But:

It is a known problem that agentic federated search through MCP isn't optimal for retrieval relevance
Agentic search is generally really slow without pre-indexing
When you interleave diverse sources of data, you need to weight each item properly
Different types of data require different query interfaces - from SQL to embedding search

Looking forward to seeing how this materializes. Our main focus today is unlocking document-based context and we'd be keen to see how this integrates with SaaS/structured data.\n\nQT @john_ssuh: Increasingly, I believe companies may need to be rebuilt from the ground up, where you have a single timeline of all observability + product metrics + file changes laid out in a retrievable system, like Datadog + Posthog + Google Drive + Slack (really unified filesystem of Claude Code chats + Codex chats). This might be the new data foundation for any and all companies to maximize AI. Needs to be rebuilt because keeping track of diffs on existing system basically impossible to produce longitudinal information on decisions and rollbacks, something coding agent storage companies are actively trying to figure out, but this should extend to businesses as a whole.

Highly skeptical existing businesses will adopt this though because it means overhauling everything about their instrumentation and business data, but I think businesses built on this foundation probably can execute 100x better and faster

See 3 related tweets

@pmarca: Interesting.\n\nQT @john_ssuh: Increasingly, I believe companies may need to be rebuilt from the gro...
@owenbjennings: https://t.co/GJprqE76ao\n\nQT @john_ssuh: Increasingly, I believe companies may need to be rebuilt f...
@ferologics: RT @john_ssuh: Increasingly, I believe companies may need to be rebuilt from the ground up, where yo...

13. business (Group Score: 128.8 | Individual: 31.3)

Cluster: 5 tweets | Engagement: 49 (Avg: 80) | Type: Tech

What does the successful launch of Elon Musk’s rockets-to-AI conglomerate SpaceX mean for the rest of the market? Which IPOs are next? And how do rich people help keep democracy healthy? All in this week’s Merryn Talks Money podcasts. https://t.co/qxU6ENaBnw

See 4 related tweets

@GuntherEagleman: 🚨 SPACEX GOES PUBLIC, ELON MUSK ABOUT TO BECOME THE FIRST TRILLIONAIRE

Huge day on Wall Street!

S...

@BusinessInsider: SpaceX's long-awaited IPO filing is set to create one of the largest wealth events in Silicon Valley...
@business: HAPPENING NOW: Follow our live blog analyzing and detailing the news from SpaceX’s first day of trad...
@business: Elon Musk has come full circle. His rocket, satellite and AI venture SpaceX hit the market in the bi...

14. sydneyrunkle (Group Score: 121.1 | Individual: 39.5)

Cluster: 5 tweets | Engagement: 63 (Avg: 25) | Type: Tech

everybody's talking about loops!! how can you instrument them with langchain?

token loop supported by a model (choose any model with langchain)
create_agent gives you the agent loop (model + tools + repeat until done)
deepagents gives you the self verification loop (agent loop + verify + repeat until satisfied)
deployments give you the meta loop (trigger agent runs in reaction to events that help improve a system)
i think the ??? loop is what we're trying to close with engine: run an agent over each trace and figure out what to tweak - prompts, tools, self verification, etc so that your meta loop is more effective per cycle.\n\nQT @latentspacepod: [AINews] Loopcraft: The Art of Stacking Loops

@RichardSSutton has his “Bitter Lesson” for models. We now have the Salty Lesson for agents:

Don’t fix things yourself, as you have done historically.

Instead focus on systems that scale with more agents, like goals and orchestration.

More in today's op-ed: https://t.co/EyYV4VGRpi

See 4 related tweets

@latentspacepod: [AINews] Loopcraft: The Art of Stacking Loops

@RichardSSutton has his “Bitter Lesson” for models. W...

@LangChain: It's easy to build loops with LangChain/DeepAgents!\n\nQT @sydneyrunkle: everybody's talking about l...
@LangChain: Great insights on loops from @sydneyrunkle\n\nQT @sydneyrunkle: everybody's talking about loops!! ho...
@Vtrivedy10: RT @sydneyrunkle: everybody's talking about loops!! how can you instrument them with langchain?

15. rohanpaul_ai (Group Score: 109.9 | Individual: 42.8)

Cluster: 5 tweets | Engagement: 109 (Avg: 43) | Type: Tech

The Information: Anthropic is reportedly moving from renting cloud compute to leasing and managing data centers itself, with more than 1GW of planned U.S. capacity and Google potentially backing the lease payments.

The old model is simple: Anthropic pays cloud providers for GPUs or custom chips, but the cloud company controls the building, power, networking, cooling, and much of the hardware schedule.

The new model puts Anthropic closer to the factory floor of AI, where data centers are not office buildings but power-heavy machine rooms built to feed training and inference systems around the clock.

To give context of its current situation, until now, Anthropic’s compute strategy has mostly run through cloud providers, with over 10GW of server rentals committed, including a $200 billion Google agreement.

The company has also locked in large cloud deals with Akamai, AWS, CoreWeave, and Fluidstack, covering Amazon’s Trainium hardware and a $50 billion Fluidstack partnership.

It has also expanded its data center team and signed a SpaceX/xAI lease for the whole Colossus 1 data center at $1.25 billion a month, plus Colossus II space.\n\nQT @theinformation: Anthropic is moving forward with a plan to control its own servers for developing AI, giving it the ability to cut its computing costs in the long run.

The AI startup has discussed arrangements under which Google would provide a financial guarantee for some of its lease obligations

Full story: https://t.co/pkkrfiZjBk

See 4 related tweets

@theinformation: Anthropic is moving forward with a plan to control its own servers for developing AI, giving it the ...
@Cointelegraph: 🚨 NEW: Anthropic is reportedly pursuing data center leases with over 1GW combined capacity and seeki...
@WesRoth: Frontier AI independence is expensive. Very expensive.

Anthropic is reportedly pursuing its first d...

@ShanuMathew93: Anthropic doesn't have any data centers so they're all leased...? lol\n\nQT @blockspace: Anthropic i...

16. Techmeme (Group Score: 108.8 | Individual: 24.0)

Cluster: 7 tweets | Engagement: 4 (Avg: 12) | Type: Tech

Google sues Chinese cybercrime network Outsider Enterprise, accusing it of using Gemini AI to create fake websites and scam hundreds of thousands of Americans (@ceciliakang / New York Times)

(Visit Techmeme dot com for the link and full context!)

See 6 related tweets

@WSJTech: Google sued scammers that it says are using the company’s Gemini AI model to create phishing website...
@Reuters: Google targets AI-powered phishing in New York lawsuit https://t.co/WMfpxu0nvG https://t.co/WMfpxu0n...
@business: Google has filed a lawsuit against a suspected Chinese cybercrime operation alleging that it used AI...
@TechCrunch: Google sues alleged Chinese cybercrime operation that used AI to send scam texts https://t.co/q61Ins...
@techdaily24: Chinese cybercrime operation that used AI to scam ‘hundreds of thousands of victims’ sued by Google ...

17. rohanpaul_ai (Group Score: 107.8 | Individual: 33.5)

Cluster: 4 tweets | Engagement: 22 (Avg: 43) | Type: Tech

Beautiful paper from Google DeepMind.

Explains the pathways from AGI to ASI, and why that jump could happen through several routes.

The authors frame the AGI-to-ASI transition around 4 technical pathways:

continued scaling of compute, model size, data, and test-time inference;
algorithmic paradigm shifts beyond today’s transformer-based foundation-model stack;
recursive self-improvement, where AI accelerates AI R&D and improves future systems; and
multi-agent collective intelligence, where large populations of specialized agents coordinate into a superhuman group agent.

Scaling may work for a while, but it could hit limits in data, compute, energy, or weaker returns from making systems larger.

Recursive improvement is the most uncertain path, because AI could speed up AI research, but that loop may also slow if hard research problems need real-world testing, scarce hardware, or new ideas.

Multi-agent collectives may be the most underappreciated path, because a society of competent digital workers could outperform a brilliant individual model through specialization, speed, and coordination.

The big point is that ASI may not arrive as 1 sudden event, but as a chain of faster changes as AI helps create better AI and stronger scientific tools.

Link – arxiv. org/abs/2606.12683

Title: "From AGI to ASI"

See 3 related tweets

@mark_k: New paper from @GoogleDeepMind: "From AGI to ASI". It argues that AGI may not be a stable endpoint. ...
@robertwiblin: RT @sebkrier: How do we go from AGI to Superintelligence? New report discusses four potential pathwa...
@Kyrannio: A great read here.\n\nQT @sebkrier: How do we go from AGI to Superintelligence? New report discusses...

18. Rajananandan (Group Score: 107.7 | Individual: 28.7)

Cluster: 5 tweets | Engagement: 24 (Avg: 60) | Type: Tech

Enormous gratitude to #IndiaAIMission for supporting the building of Sovereign models in India. Thank you @SecretaryMEITY for launching #Varya the lowest cost AI video model in the world, this morning. Varya, built by @Avataar, is built for a billion Indians @GoI_MeitY @peakxvpartners\n\nQT @GoI_MeitY: Listen to Shri S. Krishnan, Secretary, MeitY, as he highlights a significant milestone in India's AI journey with the launch of one of the foundational models supported under the IndiaAI Mission.

#IndiaAI #MeitY #DigitalIndia @SecretaryMEITY | @OfficialINDIAai https://t.co/wQFDHE3jUm

See 4 related tweets

@SarithaRai: An AI startup founded by a former Wall St. banker has launched India’s first homegrown video AI mode...
@peakxvpartners: RT @GoI_MeitY: Listen to Shri S. Krishnan, Secretary, MeitY, as he highlights a significant mileston...
@Rajananandan: .@Avataar launched #Varya today. Lowest cost video AI model in the world! This is a sneak peak of wh...
@Rajananandan: RT @OfficialINDIAai: India's AI innovation ecosystem continues to grow stronger.

With support from ...

19. VKazulkin (Group Score: 107.7 | Individual: 57.8)

Cluster: 3 tweets | Engagement: 4680 (Avg: 132) | Type: Tech

RT @karpathy: This is a super exciting release - Claude Fable 5 is the same underlying model as Mythos but with added safeguards. The benchmarks are great and it's SOTA on everything by a margin but I'll add that qualitatively also, this is a major-version-bump-deserving step change forward (imo of the same order as Claude 4.5 was in November), peaking especially for long problem-solving sessions on very difficult problems. You can give it a lot more ambitious tasks than what you're used to, the model "gets it" and it will just go, and it's never felt this tempting to stop looking at the code at all (but don't do this in prod!). The model still has quirks that people will run into and the safeguards are configured to be a little too trigger happy for launch, which can hopefully be tuned over time.

I feel a lot of things changing as working software increasingly comes out on a tap. The Jevon's paradox kicks in and I feel my own demand for software growing substantially. You can ask for anything - explainers, visualizers, dashboards, bespoke single-use apps (e.g. a full wandb that is hyper-specific just for your project), you can 10X your test suite, auto-optimize code, run giant research projects with custom HTML for the results, anything! "Free your mind" (Matrix ref). Really looking forward to all the things people build!

See 2 related tweets

@0x_kaize: FREE CLAUDE FABLE 5 and almost nobody knows this trick

Anthropic recently released their most capab...

@PaulSolt: Fable 5 can cook.

First impressions after spending ~5 hours on a semi-complex onboarding feature.

...

20. ajambrosino (Group Score: 100.9 | Individual: 47.0)

Cluster: 3 tweets | Engagement: 1864 (Avg: 420) | Type: Tech

we shipped the Tibo button\n\nQT @OpenAI: We heard you wanted to use Codex rate limit resets on your own time.

Starting today, we’re rolling out the ability to save rate limit resets to use later.

We’re starting Go, Plus, Pro, and Business users with one free reset: https://t.co/gucyTi04wc

See 2 related tweets

@thsottiaux: Heard your (amusing) feedback that it was at times annoying to receive a reset of your Codex usage w...
@kimmonismus: This is so awesome!

OpenAI is now letting Codex users save their rate limit resets and use them lat...