顶级科技推文 - 2026-07-09

2026年7月9日科技每日简报

Today's top tech conversations are led by @JamesMontemagno, whose post about 'RT @OpenAI: GPT-5.6 Sol, along...' garnered the highest engagement. Key themes trending across the top stories include https, model, voice, openai, launch. The community is actively discussing recent developments in AI, engineering practices, and startup strategies.

1. JamesMontemagno (Group Score: 1255.5 | Individual: 55.2)

Cluster: 44 tweets | Engagement: 12744 (Avg: 369) | Type: Tech

RT @OpenAI: GPT-5.6 Sol, along with Terra and Luna, will launch publicly this Thursday.

We’re expanding preview access globally now. https://t.co/Uk5HcfSc2e

See 43 related tweets

@haydenbleasel: ☀️ I've been testing GPT 5.6 Sol internally for the last few months and here is the most important t...
@SIGKITTEN: Better call Sol\n\nQT @OpenAI: GPT-5.6 Sol, along with Terra and Luna, will launch publicly this Thu...
@gdb: Sol is rising. It’s a good model.\n\nQT @OpenAI: GPT-5.6 Sol, along with Terra and Luna, will launch...
@petergyang: So what is the qualification criteria for folks to get early access to this model?\n\nQT @OpenAI: GP...
@deredleritt3r: I was an early tester of GPT-5.6 Sol.

More detailed thoughts to follow soon, but I will just confir...

2. scaling01 (Group Score: 553.4 | Individual: 47.6)

Cluster: 21 tweets | Engagement: 4170 (Avg: 333) | Type: Tech

RT @SpaceXAI: Announcing Grok 4.5, our first model trained specifically for coding and agents. It was trained with Cursor and offers frontier intelligence at leading speeds and cost efficiency.

https://t.co/i8HpU7w64k https://t.co/oBjGtTsoNc

See 20 related tweets

@kentcdodds: RT @cursor_ai: We've partnered with SpaceXAI to train Grok 4.5.

It’s our most powerful model yet an...

@cryptopunk7213: best $60 billion ever spent. the cursor acquisition needs to be studied.

grok 4.5 is only 1.4T in s...

@elonmusk: Correct\n\nQT @tetsuoai: Grok 4.5 goes public tomorrow. Here’s everything that’s known.

The model r...

@Designarena: BREAKING: Grok 4.5 by @SpaceXAI is 5th on Website Arena with an Elo of 1328.

This is a 25 rank jump...

@0xSero: Elon got his frontier grok\n\nQT @cursor_ai: We've partnered with SpaceXAI to train Grok 4.5.

It’s ...

3. shoubin621 (Group Score: 424.3 | Individual: 57.2)

Cluster: 17 tweets | Engagement: 2746 (Avg: 267) | Type: Tech

RT @OpenAI: Introducing GPT-Live, a new generation of voice models for natural human-AI interaction.

Rolling out in ChatGPT starting today.

You’ll want to turn the sound on for this one. https://t.co/WzoQFvA5ir

See 16 related tweets

@milesdeutscher: OpenAI has completely dominated AI voice and conversational prompting.

No one else even comes close...

@testingcatalog: BREAKING 🔥: OPEANAI LAUNCHED GPT LIVE 1, A NEW FULL DUPLEX VOICE MODEL FOR CHATGPT!

It comes in 2 v...

@_simonsmith: Watching the livestream, this felt like the closest I've ever seen an AI come to being a capable hum...
@reach_vb: Excited for GPT-Live-1 and mini!

It’s a pretty good model and feels shockingly natural. It’s a legi...

@OpenAIDevs: We're bringing GPT-Live-1 and GPT-Live-1 mini to the API soon.

Developers and enterprises can sign ...

4. danshipper (Group Score: 412.2 | Individual: 39.8)

Cluster: 20 tweets | Engagement: 1075 (Avg: 142) | Type: Tech

GPT-5.6 is a much better writer than Fable.

It consistently one-shots marketing emails for @tedescau that every previous model would fail at. Fable is too verbose and liable to fall into using sentences in its own private language.

If you use AI for writing, 5.6 is a fantastic model for you.

See 19 related tweets

@petergostev: My view of: Fable 5 vs GPT-5.6-Sol. They are not easy models to compare, these are my vibes - take t...
@clairevo: GPT-5.6 is like a AWD hybrid minivan, i have irrational love for her. a modern marvel. reliable. you...
@scaling01: and that's still the old Fable

Anthropic is in their own tier right now\n\nQT @mattshumer_: I had e...

@AndrewCurran_: EM had early access to GPT-5.6. A different feel is good, model diversity is good. I think it's a be...
@kieranklaassen: GPT-5.6 is very good at doing a lot of work. It's not as good as Fable but it's way faster and not t...

5. hwchase17 (Group Score: 359.8 | Individual: 52.7)

Cluster: 12 tweets | Engagement: 522 (Avg: 46) | Type: Tech

RT @PrimeIntellect: Announcing our $130M Series A to build the Open Superintelligence Stack

Led by Radical Ventures, with NVIDIA, Intel Capital, Dell Capital, and existing investors

Train, deploy, and continuously improve your own models using our stack.

Own your intelligence. https://t.co/BM31LfVUNQ

See 11 related tweets

@xeophon: Superintelligence is the most important technology of our lifetime.

We must ensure that it’s not i...

@eliebakouch: intelligence should be accessible. companies and researchers should own the models they build with. ...
@alexocheema: Every company will own and host its own models.

Congrats to everyone at @PrimeIntellect! https://t....

@latkins: I'm coming out today as a proud investor in prime intellect.\n\nQT @PrimeIntellect: Announcing our $...
@hwchase17: Congrats to prime intellect! Love partnering with them on LangChain labs work\n\nQT @PrimeIntellect:...

6. scaling01 (Group Score: 291.5 | Individual: 39.1)

Cluster: 11 tweets | Engagement: 1096 (Avg: 333) | Type: Tech

OpenAI is back baby

"GPT-6 will be based on a new, significantly larger pretrain than GPT-5.5/GPT-5.6"\n\nQT @synthwavedd: 🚨 SCOOP(s):

GPT-5.6 will be the final model in the 5.x series. GPT-6 is slated to launch in about a month, earlier than expected, and possibly even later this month
GPT-6 will be based on a new, significantly larger pretrain (versus the ~4T 5.5/5.6 'Spud' base)
There is lots of excitement at OpenAI over this new base, which they believe will be much better able to compete with both Fable 5 and upcoming 5.1, targeting a similar release window. OpenAI initially intended to continue with Spud through GPT-6, but decided against it
On the topic of Fable 5.1, it is in the late stages of the pipeline at Anthropic and a release is expected "in the coming weeks"
On the other side of the globe, DeepSeek are preparing for an imminent launch of V4 GA, which seems likely to be on par with or better than GLM-5.2, and have begun work on a new, larger model that will compete with the upcoming 2.7T MiniMax Pro

See 10 related tweets

@adonis_singh: its crazy that some people could be using gpt-6 as we speak

we will literally get gpt-6 before gta-...

@kimmonismus: If that’s true, crazy times lie ahead:

GPT-6 at the end of the month or early August, and Fable 5....

@kimmonismus: Judging by all the GPT-5.6 Sol early-access posts, it sounds as though OpenAI has achieved the major...
@AlexFinn: Tomorrow ChatGPT 5.6 sol launches. Early reviews are saying it's incredible

I've been using Fable 5...

@cryptopunk7213: massive news - we’re getting GPT 6 and Fable 5.1 in < 4 weeks, openai’s trained gpt 6 on a brand new...

7. hwchase17 (Group Score: 250.7 | Individual: 37.9)

Cluster: 8 tweets | Engagement: 130 (Avg: 46) | Type: Tech

Really excited to partner with @nvidia on the NemoClaw Deep Agents Blueprint

Deep Agents is a fully open source agent harness that we are tuning to make perform incredibly well with open models\n\nQT @LangChain: Introducing the NemoClaw Deep Agents Blueprint, a reference architecture for building open agent systems developed with @NVIDIA

✅ A fully open stack enterprises can own and customize ✅ Benchmark-leading performance ✅ Over 10x lower inference costs

Blog: https://t.co/9QumeM3897

Video: https://t.co/KmTKS8r2fk

See 7 related tweets

@LangChain: Introducing the NemoClaw Deep Agents Blueprint, a reference architecture for building open agent sys...
@nvidia: Enterprises don’t just need agents that perform. They need agents they can shape, govern and improve...
@LangChain: Excited to team up with partners across the AI infrastructure + enterprise ecosystem.

@EY_US @baset...

@LangChain: Inside the @NVIDIA NemoClaw Deep Agents Blueprint: ✅ Nemotron 3 Ultra: Open-weight model layer, cust...
@Vtrivedy10: you can take open intelligence, open source primitives for infra+security, good evals, and run loops...

8. bridgemindai (Group Score: 250.1 | Individual: 34.9)

Cluster: 9 tweets | Engagement: 2627 (Avg: 862) | Type: Tech

The craziest 48 hours in AI history just got confirmed.

TODAY: Grok 4.5 goes public.

Elon says it is Opus class, but faster, more token efficient, and lower cost.

TOMORROW: GPT 5.6 Sol, Terra, and Luna all launch.

Two frontier launches in two days.

If Grok 4.5 really is Opus 4.8 level at a fraction of the cost, SpaceXAI is finally on the map.

Both models hit BridgeBench the minute they drop.

See 8 related tweets

@XFreeze: Grok 4.5 goes live tomorrow

It is built on SpaceXAI's new V9 foundation model with 1.5 trillion par...

@astropol0: looking at the benchmark chart, Grok 4.5 is performing really well especially in Terminal Bench (83....
@nickvasiles: just configured Grok 4.5 with my hermes agent and

wow!

it is fast and feels like opus 4.7 on fas...

@coinbureau: ⚡️ JUST IN: MUSK LAUNCHES GROK 4.5, CALLING IT "OPUS-CLASS" AI MODEL

Musk just released Grok 4.5 on...

@bridgemindai: GROK 4.5 JUST WENT LIVE IN CURSOR.

256k context. Built for long running agentic coding.

Elon's cl...

9. victormustar (Group Score: 216.2 | Individual: 43.1)

Cluster: 8 tweets | Engagement: 298 (Avg: 57) | Type: Tech

RT @robbyant_brain: Today we open-source LingBot-Video — the first MoE-based video foundation model built for embodied intelligence. 🔹30B params, only 3B active at inference. 🔹Augmented with 70K hours of embodied data on top of large-scale internet video pretraining. 🔹Already outperforming Wan2.6, Seedance 1.5 Pro, and Cosmos3 Super on RBench. 🧵👇

See 7 related tweets

@omarsar0: Massive open-source release!

LingBot-Video isn't about video quality; it optimizes for physical rea...

@heyshrutimishra: New video model just dropped. But this one isn't built for cinematic video.

LingBot-Video is desi...

@svpino: An open-source foundation model for robots.

Show me something more hardcore than this, and I'll dea...

@ivanfioravanti: RT @_akhaliq: LingBot-Video is out on Hugging Face

MoE-based video foundation model built for emb...

@DataChaz: RT @DataChaz: GREAT NEWS FOR THE ROBOTICS SPACE

🚨 @Robbyant_brain just open-sourced LingBot-VLA 2.0...

10. btibor91 (Group Score: 174.9 | Individual: 37.0)

Cluster: 7 tweets | Engagement: 222 (Avg: 166) | Type: Tech

OpenAI launched GPT-Live, a new generation of full-duplex voice models (GPT-Live-1 as the default for Go, Plus and Pro users, GPT-Live-1 mini for Free) that can listen and speak at the same time, delegates deeper work to GPT-5.5 behind the scenes and is rolling out to ChatGPT users globally today across iOS, Android and ChatGPT[.]com, with API access planned soon

GPT-Live continuously processes input while generating output, making interaction decisions many times per second (whether to speak, continue listening, pause, interrupt or invoke a tool), and delegates questions requiring search, reasoning or more agentic capabilities to another model like GPT-5.5 as it keeps the conversation going, with GPT-Live-1 (instant) and the mini using GPT-5.5 Instant in the background and GPT-Live-1 Medium and High using GPT-5.5 Thinking with medium and high reasoning effort
The new ChatGPT Voice acknowledges what you say with phrases like mhmm or got it, waits instead of jumping in when you pause to think, focuses better on your voice in background noise, shows rich visual cards for topics like weather, stocks and sports, keeps supporting search, memory, images and file uploads, comes with nine remastered voices and an Intelligence setting with Instant, Medium and High levels
At launch GPT-Live does not support voice with video or screen sharing (those remain in Advanced Voice Mode), and Live is not initially available in Temporary Chats, the desktop app, Work, Codex, custom GPTs or Business, Enterprise and Edu workspaces

I have been testing GPT-Live-1 in ChatGPT for a while now, and it is much better, more intelligent and more natural than the previous Advanced Voice Mode, plus more fun to use and more useful

See 6 related tweets

@testingcatalog: OPENAI 🔥: GPT‑Live‑1 offers significant improvements over AVM across almost all dimensions.

GPQA...

@testingcatalog: A new GPT Live 1 voice mode is rolling out super fast. It also comes with an effort selector and upd...
@Techmeme: OpenAI launches GPT-Live, a new generation of voice models built on a full-duplex architecture, mean...
@adonis_singh: we got rl-maxxing on voice models now😭\n\nQT @Angaisb_: OpenAI introduces GPT-Live, a new generation...
@MTSlive: SITUATION DETECTED: OpenAI has launched GPT-Live, a new generation of voice models now powering Chat...

11. victormustar (Group Score: 174.3 | Individual: 37.2)

Cluster: 7 tweets | Engagement: 40 (Avg: 57) | Type: Tech

I swear it would have been a better business move to open source it 🤷‍♂️\n\nQT @AIatMeta: Introducing Muse Image and Muse Video, the first media generation models developed by Meta Superintelligence Labs.

Muse Image is our most advanced image generation model yet. It follows instructions faithfully, edits with precision, composes from multiple references, and draws on Instagram for social context. It also brings agentic tool use capabilities to image generation and integrates with Muse Spark.

You can try Muse Image in the Meta AI app and web, as well as in Instagram Stories and WhatsApp – starting in limited countries with more locations on the way.

Today we’re also previewing Muse Video, which is built upon the same pretraining base as Muse Image to deliver exceptional visual fidelity with native audio support.

Learn more about both models: https://t.co/QtKDPDZP5v

See 6 related tweets

@IamEmily2050: Congratulations to everyone at Meta and the MSL team for finally taking positive steps into image an...
@TeksEdge: Good work Meta! We've been waiting for some good news from the AI team. Your new Muse Image model h...
@gmi_cloud: Highly impressed by Muse Image! It beats Nano Banana 2 across 7 tests, tho GPT Image 2 is still the ...
@WesRoth: Meta is launching Muse Image, its first image generation model from Meta Superintelligence Labs, now...
@AngryTomtweets: RT @AngryTomtweets: Meta Superintelligence Lab just dropped a new image model called Muse Image.

A ...

12. 0xSero (Group Score: 172.1 | Individual: 38.1)

Cluster: 9 tweets | Engagement: 266 (Avg: 162) | Type: Tech

Losing access to Chinese Open Weight models would put tremendous pressure on Nvidia's open source team.

Even Nvidia's used Chinese Open Weight models for their own developments.

Why go down this road? Isn't the world already closed enough as it is?

https://t.co/Z6WpNfSf7T\n\nQT @jukan05: CHINA CONSIDERS RESTRICTING OVERSEAS ACCESS TO CUTTING-EDGE AI MODELS

China’s Ministry of Commerce has led meetings over the past month with major AI companies, including Alibaba, ByteDance, and https://t.co/YDe0KRldDB, to discuss measures that would restrict overseas access to cutting-edge AI models, including models that have not yet been released.

The discussions reportedly include not only closed-source models but also open-weight models. However, the scope of application is still under debate, and the rules may ultimately apply only to future frontier models.

Officials have also discussed designating the leakage or theft of proprietary AI technologies as a national security crime, with stronger penalties, as well as restricting the types of foreign capital that can invest in Chinese AI startups.

The backdrop is the U.S. move to strengthen export controls on AI models, along with national security concerns over cutting-edge models that could possess advanced cyberattack capabilities.

Chinese authorities are reportedly concerned that advanced U.S. cybersecurity AI models could be used to exploit vulnerabilities in Chinese software.

Since the beginning of this year, China has continued to tighten measures to prevent AI technology from being transferred overseas. Authorities have investigated whether Chinese AI startups that relocated abroad violated export control laws, while also strengthening oversight of overseas transactions involving Chinese investors, technology, data, and national security concerns.

Future regulations could take the form of a tiered framework based on technological capability. Basic open-source AI models may be managed through a filing system, high-performance models may be subject to security reviews, and the most sensitive frontier models may be banned from public release or restricted to use within China.

See 8 related tweets

@rohanpaul_ai: RT @rohanpaul_ai: 🇨🇳This is bad news. Now, frontier open source access also will probably no more be...
@ChrisRMcGuire: Key excerpt: China's own AI companies are still telling the Chinese government "they lack enough hig...
@Forbes: Chinese authorities are in talks to potentially curb overseas access to the country’s leading AI mod...
@Cointelegraph: 🇨🇳 JUST IN: China is considering curbing overseas access to its top AI models, including unreleased ...
@WesRoth: China is reportedly considering restrictions on overseas access to its most advanced AI models.

Off...

13. teortaxesTex (Group Score: 170.9 | Individual: 31.1)

Cluster: 6 tweets | Engagement: 67 (Avg: 59) | Type: Tech

➤ Pricing of $2/$ 6 per 1M tokens of input/output; cache hits are discounted by 75% to $0.5 per 1M tokens, and costs still double with long (>200k token) inputs if only they had V4 arch…\n\nQT @ArtificialAnlys: SpaceXAI’s Grok 4.5 scores 54 to place fourth on the Artificial Analysis Intelligence Index following only Fable 5, GPT-5.5, and Opus 4.8. It scores on par with GPT-5.5 in Codex on the Artificial Analysis Coding Agent Index in the Grok Build harness, at much lower cost

Grok 4.5 improves 16 points over Grok 4.3 on the Intelligence Index, bringing SpaceXAI to the intelligence frontier behind only OpenAI and Anthropic, and outperforming all open weights models and notably Google’s Gemini models. Key standout areas of performance are agentic knowledge work and coding.

Grok 4.5 in Grok Build scores 76 on the Artificial Analysis Coding Agent Index, on par with GPT-5.5 (xhigh) in Codex and just below Fable 5 (max) in Claude Code, and at a small fraction of the token usage and price.

Congratulations to @SpaceXAI, @cursor_ai, and @elonmusk on the impressive release!

Key Takeaways:

➤ Grok 4.5 performs very strongly on agentic tasks. Grok 4.5 ranks #4 on GDPval-AA v2 with an Elo of 1543, between Claude Opus 4.8 (1600) and GLM-5.2 (1513). It achieves the top score on 𝜏³-Banking of 33%, above 31% from GPT-5.5 (xhigh), and sits on the cost vs performance Pareto frontier across all three agentic evaluations in the Intelligence Index ➤ Grok 4.5 is one of the most cost efficient models to run for near-frontier intelligence. It costs $0.31 per task on the Artificial Analysis Intelligence Index and$ 2.59 per task on the Artificial Analysis Coding Agent Index within Grok Build ➤ Low cost for Grok 4.5 is driven by both low pricing and token efficiency. Grok 4.5 has a headline price over 60% lower than Claude Opus 4.8 and GPT-5.5, and used ~14k output tokens per Intelligence Index Task - over 60% lower than Opus 4.8. On the Coding Agent Index, Grok 4.5 stands out on the Pareto frontier of Coding Agent Index score vs. Total Tokens, using only 1.9M tokens for the Coding Agent Index while scoring 76 ➤ As a coding agent, Grok 4.5 in Grok Build is on par with GPT-5.5 and offers efficiency benefits: In our Artificial Intelligence Coding Agent Index that consists of DeepSWE, Terminal-Bench v2, and SWE-Atlas QnA, Grok 4.5 in Grok Build ranks third, on par with GPT-5.5 (Codex) and below Fable 5 (Claude Code). It is also very efficient in achieving this result: Grok 4.5 in Grok Build cost $2.49 per task while Fable 5 in Claude Code cost$ 11.80 and GPT-5.5 in Codex $5.07. This is driven by relatively low token pricing and the model using far fewer tokens than comparable models (1.9M average tokens used per task), significantly less than Fable 5 in Claude Code (7.2M) and GPT-5.5 in Codex (6.2M)

Other model details:

➤ Context window of 500k tokens - a reduction from Grok 4.3’s 1M token context, but retaining configurable reasoning and vision input

➤ Pricing of $2/$ 6 per 1M tokens of input/output; cache hits are discounted by 75% to $0.5 per 1M tokens, and costs still double with long (>200k token) inputs

➤ As Elon Musk has disclosed, Grok 4.5 is 3x larger than its predecessor at 1.5T parameters

See 5 related tweets

@IBuzovskyi: HERMES AGENT NOW RUNS GROK 4.5. 1.5 TRILLION PARAMETERS. TRAINED ALONGSIDE CURSOR ON REAL DEVELOPER ...
@BrianRoemmele: I find that @Gork 4.5 not only exceeds Claude and ChatGPT in most cases, it has a far deeper and wid...
@elonmusk: Grok 4.5 context window will upgrade to 1M probably by next week\n\nQT @ArtificialAnlys: SpaceXAI’s ...
@BrianRoemmele: BOOM GROK 4.5 IS OUT!

Grok 4.5 has now been officially released and is available via the SpaceXAI A...

@zerohedge: RT @ArtificialAnlys: SpaceXAI’s Grok 4.5 scores 54 to place fourth on the Artificial Analysis Intell...

14. MTSlive (Group Score: 167.8 | Individual: 36.2)

Cluster: 5 tweets | Engagement: 71 (Avg: 61) | Type: Tech

DAILY SITUATION RECAP (via @theojaffee):

Chinese AI labs are refocusing on domestic chips. Chinese companies currently rely on a mix of smuggled Nvidia chips, legal export-controlled versions of American chips like Nvidia’s H20 and A800 and AMD’s MI308, and Chinese chips like the Huawei Ascend, Hygon DCU, and Cambricon MLU/Siyuan. Due to tensions with the US and mandates (and funding) from the Chinese government to encourage the domestic chip ecosystem, Chinese companies are reallocating their budgets toward local chips.

DeepSeek and ZAI will make their own chips. AI demand continues to skyrocket, and compute capacity continues to be hard to come by, especially in China where US chips are export-controlled. Labs also like to have their own chips in order to reduce dependence on external compute suppliers (like Nvidia or Huawei) and make optimizations for their specific stack. OpenAI just announced their Jalapeño chip in collaboration with Broadcom, and Anthropic is rumored to be working on a chip with Samsung.

Meta releases Muse Image and announces Muse Video, the first image and video model from Meta Superintelligence Labs. They seem like strong models — Muse Image ranks second on Arena’s leaderboard behind just GPT Images 2.0, and Muse Video is behind only Gemini Omni Flash and Seedance 2.0. Teortaxes points out that the underlying model is strong enough to replicate a QR code without information loss. MSL’s most recent language model release was Muse Spark three months ago.

Claude Fable access in paid Claude plans is extended to July 12. Claude Pro, Max, and Team users can use Fable 5 as part of their plan up to 50% usage until end of day Sunday. This promotion was supposed to end today (July 7). Afterwards, users will have to pay $10/$ 50 per million input/output tokens.

Mercor hits $2 billion ARR. The company pays contractors with specific domain knowledge to create specific training data for frontier labs, supplementing the general training corpus of the Internet. 60-70% of Mercor’s revenue is paid out to contractors, and the company is cash flow positive.

Four US states are suing Meta for $1.4 trillion. Yes, trillion with a T. Colorado, California, Kentucky, and New Jersey allege that the company deliberately designed its products to addict minors and misled the public about their safety. Populist and government backlash against both real and perceived abuses of Big Tech continues to grow.

UN Secretary-General Antonio Guterres opposes autonomous weapons. He called for them to be “banned by international law”, labeled them “morally repugnant”, and insisted that “some decisions must remain forever human.” In the long run, however, militaries that refuse the adoption of autonomous weapons for moral reasons may be at a permanent structural disadvantage.

Samsung begins mass producing storage for Nvidia Vera Rubin. Vera Rubin, Nvidia’s latest hardware generation after Blackwell, is undergoing mass production and is scheduled to release within months. Samsung has begun production of the PM1763 enterprise solid-state drive to provide storage for Vera Rubin datacenter systems.

Amazon is selling $25B in bonds in order to help fund its hyperscale AI infrastructure buildout.

Microsoft is replacing OpenAI and Anthropic models with Microsoft AI models in some products like Excel and Outlook in order to cut costs.

See 4 related tweets

@jukan05: Some thoughts I’ve been having recently:

Could DeepSeek and Zhipu considering their own chips imply...

@ahall_research: These recaps are becoming indispensable, seem to be getting better and better\n\nQT @MTSlive: DAILY ...
@KyleReidhead: China plans to let top AI firms including Alibaba, ByteDance and DeepSeek buy limited Nvidia H200 ch...
@chris_j_paxton: RT @ruima: With the news today of DeepSeek & Zhipu both working / thinking about making their own ch...

15. willccbb (Group Score: 148.9 | Individual: 52.3)

Cluster: 6 tweets | Engagement: 1819 (Avg: 142) | Type: Tech

we raised $130M @$ 1B for our series A

>$100M run rate

we’re just getting started\n\nQT @TechCrunch: Prime Intellect raises $130M Series A to help enterprises build their own AI agents https://t.co/WWTfrjipol

See 5 related tweets

@TechCrunch: Prime Intellect raises $130M Series A to help enterprises build their own AI agents https://t.co/WWT...
@PrimeIntellect: RT @dr_alphalyrae: Prime Intellect is building the compute infrastructure and software tools compani...
@Prince_Canuma: Super happy and proud of the @PrimeIntellect team 🚀\n\nQT @TechCrunch: Prime Intellect raises $130M ...
@Techmeme: Prime Intellect, which helps companies build their own AI agents by offering computing power and spe...
@willccbb: RT @anravich94: We raised $130M @$ 1B for our series A. We have a >$100M run rate and we’re just ...

16. elonmusk (Group Score: 148.5 | Individual: 29.7)

Cluster: 6 tweets | Engagement: 5750 (Avg: 9128) | Type: Tech

Grok groks engineering.

Next month’s release will be another step-change improvement, as we close the loop on solving real-world engineering problems at Tesla, SpaceX, Neuralink and Boring Company.\n\nQT @SpaceXAI: Grok 4.5 is built for real-world engineering. It excels in large codebases and handles long-running tasks that span multiple repositories, hundreds of skills, and a variety of tools. https://t.co/70qNssFqEz

See 5 related tweets

@no_stp_on_snek: hey that's pretty good for deep swe. and already available?!?!

time to test https://t.co/qy0526cOVY...

@ivanfioravanti: It seems powerful!!!\n\nQT @cursor_ai: We've partnered with SpaceXAI to train Grok 4.5.

It’s our mo...

@Dan_Jeffries1: Sol and Grok. Tomorrow is going to be a good day.\n\nQT @cursor_ai: We've partnered with SpaceXAI to...
@elonmusk: RT @SpaceXAI: Grok 4.5 is built for real-world engineering. It excels in large codebases and handles...
@SawyerMerritt: RT @elonmusk: Grok groks engineering.

Next month’s release will be another step-change improvement...

17. wallstengine (Group Score: 146.7 | Individual: 18.9)

Cluster: 10 tweets | Engagement: 155 (Avg: 127) | Type: Tech

China plans to let top AI firms including Alibaba, ByteDance and DeepSeek buy limited Nvidia H200 chips, potentially under 200,000 total.

The $NVDA chips would be restricted to training on public data, with domestic processors prioritized for inference.

Source: The Information https://t.co/aCh6LwOB35

See 9 related tweets

@StockSavvyShay: China reportedly plans to let top AI firms buy a limited number of $NVDA H200 chips.

The approval w...

@MilkRoadAI: RT @KyleReidhead: China plans to let top AI firms including Alibaba, ByteDance and DeepSeek buy limi...
@financialjuice: China plans to let top AI firms buy a limited amount of NVIDIA H200 chips - The Information. $NVDA h...
@WhaleInsider: JUST IN: 🇨🇳 China to let top AI firms buy limited $NVDA H200 chips. https://t.co/SvWkznFEtY...
@negligible_cap: *CHINA PLANS TO LET TOP AI FIRMS BUY LIMITED AMOUNT OF NVIDIA H200 CHIPS - THE INFORMATION

Will be ...

18. matei_zaharia (Group Score: 137.2 | Individual: 30.0)

Cluster: 6 tweets | Engagement: 16 (Avg: 43) | Type: Tech

One reason we built custom coding and data agent benchmarks internally at Databricks (e.g. https://t.co/SdLbEP9tye). Academic benchmarks are great and people will build better ones, but you also care about YOUR tasks, which are often different. Each company needs its own "loop".\n\nQT @OpenAI: We audited SWE-Bench Pro, one of the most widely used AI coding benchmarks, and found it no longer reliably measures frontier coding capability.

We find 30% of SWE-Bench Pro tasks to be broken, and are retracting our previous recommendation that the research community use it as a leading coding eval. https://t.co/wDdSEjBe4F

See 5 related tweets

@Yuchenj_UW: At Databricks, we built an internal coding benchmark on our multi-million-line codebase.

Biggest ta...

@databricks: RT @matei_zaharia: We benchmarked coding agents on our own internal tasks at Databricks and learned ...
@MatthewBerman: This is pretty serious, right?\n\nQT @OpenAI: We audited SWE-Bench Pro, one of the most widely used ...
@xeophon: openai looked at the data and it stared right back :(\n\nQT @OpenAI: We audited SWE-Bench Pro, one o...
@AndrewCurran_: RT @OpenAI: We audited SWE-Bench Pro, one of the most widely used AI coding benchmarks, and found it...

19. zephyr_z9 (Group Score: 136.7 | Individual: 37.9)

Cluster: 4 tweets | Engagement: 283 (Avg: 168) | Type: Tech

Bad behavior from Anthropic\n\nQT @evanreiser: .@AnthropicAI has filed a lawsuit against @Abnormal claiming we copied their brand to mislead security customers, and they are asking for “all revenues, earnings and profits”

This is obviously not true. They don’t own every A/slash design in AI cybersecurity, and they don’t get to turn a logo dispute into a claim on Abnormal’s entire business.

It would be easier to concede quietly. But Abnormal was built on trust, innovation, and intellectual honesty.

Values are what you do when tested (especially when inconvenient) Our behavioral security platform and our AI models are ours.

Everything meaningful about Abnormal was built the hard way: the technology, the customer relationships, and the trust. We earned that trust by protecting customers, and we will defend it.

My blog post about it (link below):

See 3 related tweets

@Mayhem4Markets: Anthropic sued one of its biggest customers last week.

The CEO found out from a reporter. Not from ...

@irl_danB: guy who subscribed to Claude thinking it was an Abnormal product this whole time\n\nQT @evanreiser: ...
@alex: RT @evanreiser: .@AnthropicAI has filed a lawsuit against @Abnormal claiming we copied their brand t...

20. steipete (Group Score: 133.6 | Individual: 30.1)

Cluster: 6 tweets | Engagement: 946 (Avg: 516) | Type: Tech

Gosh I can't wait for all of you to play with this.\n\nQT @thdxr: i've never hyped a model release, we're generally conservative with how we use these things

but gpt-5.6 has had a massive impact on our team, we're using 5x the tokens as we used to

it's not even smarter than fable or anything, but it's just so reliable and fun to use

See 5 related tweets

@thdxr: one thing i am worried about is that it's possible that 50% of our enjoyment came from its speed

an...

@Teknium: I like a model that's super pleasant in Hermes Agent 👀 https://t.co/syAmcjbNII\n\nQT @theo: Apparent...
@PaulSolt: RT @robinebers: GPT 5.6 Sol/Terra/Luna is... fantastic!

some early/high-level observations from usi...

@zephyr_z9: RT @thdxr: i've never hyped a model release, we're generally conservative with how we use these thin...
@dair_ai: RT @omarsar0: Claude power users: "Fable 5 is the best"

Codex power users: "GPT-5.6 is the best"

R...