热门科技推文 — 2026年7月18日

今日科技动态：中国的月之暗面凭借 Kimi K3 成为舆论焦点。据报道，这款拥有 2.8 万亿参数、支持百万级上下文的多模态模型在前端编程基准测试中名列前茅，并对 OpenAI 和 Anthropic 的估值及技术领先地位构成挑战。该模型发布之际，中国正推动成立一个跨国人工智能治理联盟，进一步加剧地缘政治竞争。与此同时，数据中心成本飙升引发了外界对人工智能基础设施经济性的质疑；ChatGPT 更新了桌面端体验；Databricks 获得新一轮融资，但推迟了首次公开募股预期；人工智能智能体则继续探索股票交易。

1. deredleritt3r (Group Score: 333.0 | Individual: 38.6)

Cluster: 13 tweets | Engagement: 915 (Avg: 290) | Type: Tech

A few thoughts on Kimi K3:

The Moonshot team is absolutely cracked. Some of you might remember that K1.5 was released on January 20, 2025 - i.e., on the same day as DeepSeek R1, and it was ~just as good as R1. IIRC, the model wasn't open-source, which is why no one ever talks about it and DeepSeek got all the glory.
Do K3's capabilities generalize outside of coding, or is it coding-benchmaxxed? In my initial testing, it's materially worse than GPT-5.6 Sol and Fable 5 for certain use cases that do not involve coding.
I will not tire of saying this: the publicly available models in July 2026 do not matter. It doesn't matter whether China is 2 months behind the frontier, or 4 months, or 8 months. The only thing that matters is the race to RSI.

What do we need for RSI? IMO, two things: (i) a model with excellent research taste; and (ii) tons of compute.

Does excellent research taste develop on its own from increasingly strong coding abilities? I'm skeptical. I think it's a much more general ability that is more a function of overall model intelligence than just proficiency at some skill (like coding).

All that said, my estimate for how far behind China is has not changed at all with the release of K3 (IMO, much farther behind than the timeline thinks today).
This model's cyber capabilities are currently unknown, and this is an area worth watching. The few weeks following the release of K3's weights might be interesting times for defenders (or, then again, maybe not).
The geopolitical consequences of the release of this model are also worth watching - both in the U.S. and in China.

See 12 related tweets

@kimmonismus: The signs that China is catching up and closing the gap between Western closed-source and Eastern op...
@firstadopter: Kimi K3 is bullish for AI fundamentals, not bearish.

China's Kimi K3 AI Model May Trigger a DeepSee...

@scaling01: come on, show some balls make your predictions

not like Nathan tweeting something along the lines: ...

@MTSlive: .@deredleritt3r on why Kimi K3 is the best open source model but still not close to the frontier:

“...

@tenobrus: based on ryan's tests (effectively "how much mathematical reasoning can the model perform in one for...

2. shiri_shh (Group Score: 301.0 | Individual: 44.0)

Cluster: 10 tweets | Engagement: 2010 (Avg: 314) | Type: Tech

Why tf are OpenAI and Anthropic valued at hundreds of billions while Kimi is sitting at just $20B? https://t.co/NgwNdI8M7w\n\nQT @arena: Big news: Kimi-K3 by @Kimi_Moonshot is now #1 in the Frontend Code Arena with 1679 pts, surpassing Claude Fable 5.

This is a 17-place jump from Kimi-k2.6 (#18 -> #1).

In Frontend, Kimi-K3 ranked #1 in 6 of 7 domains: Brand & Marketing, Reference-Based Design, Data & Analytics, Consumer Product, Simulations, and Content Creation Tools, landing #2 only in Gaming behind Fable 5.

The full model weights will be released by July 27.

Congrats to the @Kimi_Moonshot team on this major milestone!

See 9 related tweets

@zerohedge: China is now leading the AI arms race after spending a tenth what the US did https://t.co/sJOwU0KZhp...
@zephyr_z9: "Kimi 3 distilled Fable and therefore big training infra still required for leading edge." Fable was...
@aakashgupta: Eighteen months ago DeepSeek nearly killed Moonshot AI. Kimi fell from the #3 chatbot in China to #7...
@akshay_pachaar: seven weeks. that's all it took.
BrowseComp: 84.3
DeepSWE: 59.0
GPQA Diamond: 91.0
Frontier...
@MTSlive: SITUATION ANALYSIS: Kimi K3

After weeks of teasing, Moonshot AI has finally released its long-await...

3. SiliconFlowAI (Group Score: 275.1 | Individual: 45.6)

Cluster: 9 tweets | Engagement: 13050 (Avg: 2176) | Type: Tech

RT @Kimi_Moonshot: Introducing Kimi K3: Open Frontier Intelligence

🔹 2.8 Trillion Parameters, 1 Million Context, Native Multimodal 🔹 Kimi Delta Attention enables up to 6.3x faster decoding in million-token contexts 🔹 Attention Residuals deliver ~25% higher training efficiency at <2% additional cost 🔹 Built for long-horizon agentic coding and self-evolving workflows

Kimi K3 is now live on on https://t.co/zrk6zZxZUo, Kimi Work, Kimi Code, and the Kimi API. Open Weights by July 27, 2026.

🔗 API: https://t.co/XCrgjXAqMw 🔗 Tech blog: https://t.co/YTfiMSNM1f

See 8 related tweets

@ModelScope2022: One of the most anticipated open-weight releases is coming soon 🚀 ModelScope will provide Day 0 supp...
@heynavtoor: 2.8 trillion parameters. 1 million token context. 6.3x faster decoding. native multimodal.

Kimi K3 ...

@sairahul1: Sir, even Chinese models are better than Opus 4.8 https://t.co/VcObf00TBs\n\nQT @Kimi_Moonshot: Intr...
@DataChaz: Let me show you how to run Kimi-K3 locally, right from the comfort of your home https://t.co/9mq7jCC...
@mervenoyann: et voila!\n\nQT @Kimi_Moonshot: Introducing Kimi K3: Open Frontier Intelligence

🔹 2.8 Trillion Para...

4. ml_angelopoulos (Group Score: 270.5 | Individual: 46.9)

Cluster: 12 tweets | Engagement: 3404 (Avg: 155) | Type: Tech

RT @DavidSacks: This is concerning. For the first time, a Chinese model Kimi K3 has taken #1 on the Frontend Code Arena and is scoring at or near the frontier on other benchmarks.

Meanwhile America is tying itself in knots: politicians and bureaucrats are banning new data centers, piling on state regulations, and pushing for new federal agencies to pre-approve frontier models.

This is how you lose the AI race. The rest of the world won’t play by our rules if we bog ourselves down. Permissionless innovation is how America won the internet and became the technological envy of the world. We can do it again with AI -- while addressing risks in a targeted way -- or we’ll watch our lead evaporate.

See 11 related tweets

@DavidSacks: This is concerning. For the first time, a Chinese model Kimi K3 has taken #1 on the Frontend Code Ar...
@kimmonismus: The result of the Kimi K3 release will be: less regulation, more capital expenditure, faster release...
@scaling01: guess who is advocating for acceleration?
our former "AI czar"

A lot of people on twitter got on...

@synthwavedd: Interesting tone by David Sacks re Kimi K3 here. For reference, Sacks is one of Trump's top advisors...
@vkhosla: Agree, 100% we shouldn’t be tying ourselves in knots. Even bigger issue is the brilliant talent we a...

5. deanwball (Group Score: 261.2 | Individual: 47.5)

Cluster: 8 tweets | Engagement: 1211 (Avg: 284) | Type: Tech

Some observations on Kimi:

It's a very good model! I don't think its performance can be explained away by distillation or anything like that. In agentic coding sessions, it seems pretty much on par with the best public models of Q1 2026. In my fairly limited use, it also seemed very token hungry. It's not obvious to me that this model is actually that cheap to run.
I am personally surprised the Chinese state continues to allow the open sourcing of models this good, given potential risks. To be clear, I myself might be fine with models presenting this level of marginal risk being open weight, but I am surprised that China is fine with it. I suspect the reason they are is 75% explained by strategic blindness/lack of AGI-pilledness (the CCP is very Yann Lecun-y in its views of AI). The other 25% or so is their lack of compute for customer inference (making China's open-weight strategy an unintended byproduct of US export controls) and the normal Chinese strategy of aggressive exports. For the companies, as opposed to the government, the decision to open source is partially ideological and partially because they are behind, and they know that very few people would pay for sub-frontier models from China.
Open-weight models are inherently decelerationist, and I'm continually surprised to see the so-called "accelerationists" so excited about open-weight models. I suspect the reason they are is that they know open-weight models are effectively ungovernable, and they simply like the overall cloak of ungovernability open-weight models create over the whole of AI. It's not a bad strategy; it reminds me of James Scott's recounting of the hill people in "the art of not being governed." Still, in the end, open-weight models deter further AI capex.
One probable outcome of an open-weight-model-dominant world is full AI communism, which is precisely what China proposes: rather than a market product, AI is a "public good" which will ultimately be provided by the state as a kind of "digital public infrastructure." This future strikes me as a dystopian hellscape, but I've never met an open-weight models advocate who doesn't ultimately concede this is where things end. You'd be surprised how many 'accelerationists' lobbied me, while I was in government, to support an eleven or twelve-figure federally funded data center so that startups could train models at a subsidy and then give them away for free. There was no other way for AI to progress, they said. Perhaps this is the logical end state of things. Nonetheless, I find myself surprised to see supposed accelerationists excited about such an outcome. I think many of them just don't know what they're doing. Many accelerationists do not view the creation and serving of frontier models as a legitimate business.
I would guess that the Trump Administration will at some point realize that their best strategy here would be to create large amounts of regulatory risk around the use of open-weight Chinese models. You don't need to "ban open source" (one of the dumber motifs of AI policy discussion). You just need to direct every agency to issue soft law that creates FUD. "A Federal Reserve Advisory Bulletin found that there may be backdoors in Chinese AI models." It needn't be that well justified. You just create enough regulatory risk that every regulated enterprise backs off. You probably don't want to create so much regulatory risk that you scare off the hyperscalers from serving Chinese models; this will just drive startups to sketchier providers. There's a happy middle ground here. I'd assume they will do some version of this.
It's probably true that open-weight models of this capability make the world a bit more dangerous, but not so much more that you'll really notice. At some point the models will be capable enough that you will notice. "A nonliving, invisible, dangerous, and infinitely self-replicating agent escaped from a Chinese lab," you say? Color me shocked.

See 7 related tweets

@quxiaoyin: What Kimi K3 means for USA AI:

FYI Kimi K3 is open weight and will be released on July 27, 2026....

@emollick: So I guess it is time to wonder: how does pre-clearance work for open weights models? No model card ...
@perrymetzger: Now that Chinese AIs are achieving parity with the best US created models, a few short-term predicti...
@BlancheMinerva: It continues to be an excellent heuristic that when employees of trillion dollar corporations shake ...
@cryptopunk7213: we’ve got to stop yelling “distillation attack” and realise china’s building frontier intelligence f...

6. JLopas (Group Score: 242.8 | Individual: 35.4)

Cluster: 9 tweets | Engagement: 1494 (Avg: 293) | Type: Tech

RT @GavinSBaker: Kimi K3 may be an important inflection point for AI. Potentially negative for Anthropic and OpenAI while being net positive for essentially every other company in the world. I mean that very literally. Although the real “Sputnik moment” would be an open-source frontier model that was also token efficient unlike Kimi K3 which is 50-70% more expensive to run than GPT 5.6 per Artificial Analysis.

Rationale: A world where there are only 2-3 dominant frontier labs with 90% inference margins is net negative for every other layer while being awesome for those 2-3 labs. Those labs would become monopsonies for power, data centers, semiconductors and hyperscalers and would obviously vertically integrate over time into all those layers while also completely subsuming the application/software layers. Anything that lowers margins and increases competition at the model layer is good for every other AI layer: power, semiconductors, hyperscalers, neoclouds and yes even software. This is why Jensen is so supportive of open-source. An open-source model requires the exact same amount of compute to run as a closed frontier model of similar size and architecture. Kimi K3 is roughly the same price as GPT 5.6 Terra on a per token basis, which actually suggests that it is less computationally efficient as I am sure that GPT 5.6 is priced to a higher margin than K3. And given that K3 is a token wastrel, i.e. token inefficient, it is significantly more expensive per task than GPT 5.6 and Grok 4.5, which are much more token efficient. Cost per token and token efficiency (i.e. intelligence density per token) are the drivers of intelligence per unit of cost. The winning AI companies will be those that offer the most intelligence per $over time. Lower margin % at the model layer = more margin$ at every part of the infrastructure layer and is a godsend for software. This can happen either through open-source models like K3 at the frontier or having a vertically integrated model company like Meta, SpaceX or Google at the frontier. Both outcomes result in a lower margin % at the model layer as vertically integrated model companies don’t really care where the margin $ come from. This is why it was so painful for OpenAI and Anthropic when Google was right there with them from a model competitiveness perspective and why Grok 4.5 and Muse 1.1 were just as important as Kimi K3.  The reason Kimi K3 is only potentially negative for Anthropic and OpenAI is 1) the @ericvishria point that the Claude and ChatGPT products and harnesses may be more important than their models today and 2) the hypothesis that they have much more advanced model checkpoints internally that are already being used for RSI. In the latter scenario, reaching RSI even a few months ahead of other labs might be enough to cement a permanent lead.

Time will tell on both points. And likely fairly quickly.

Caveat would be that since Kimi K3 is not token efficient and thereby actually more expensive than ChatGPT 5.6, we may need to see a more token efficient open-source model at the frontier or see Grok 5/Composer 4/Muse 2 at multiple points on the Pareto frontier for this potential risk to Anthropic and OpenAI to play out. And I am sure they will both vertically integrate as quickly as possible while continuing the product/harness strength they have shown over the last 8 months.

See 8 related tweets

@ShanuMathew93: Great insights. Gavin nails it.

If an oligopoloy of labs sustain 90% inference margins, they captur...

@benitoz: Open source just walked into the frontier club without paying the cover.

Kimi K3 is an open-weights...

@chamath: Gavin is right. This is positive for everyone except the closed frontier labs.\n\nQT @GavinSBaker: K...
@levie: This post is key. The cheaper AI gets, the more opportunity there is for the entire ecosystem - esp...
@MTSlive: Gavin Baker says Kimi K3 could be a "Sputnik moment," bad for Frontier Labs, great for everyone else...

7. teortaxesTex (Group Score: 239.5 | Individual: 39.0)

Cluster: 10 tweets | Engagement: 243 (Avg: 64) | Type: Tech

Xi's speech can also be understood as the complete victory of the Industrial Party. Now, AI is also an aspect of China's "universal value". You've got to admit, it goes hard. These are billions of people who were en route to historical irrelevance. They don't like the West much. https://t.co/OY1zA2qKGH\n\nQT @tenobrus: it looks like this might be the answer to my question today. maybe this really is about "soft power", in the sense that china is seeing america positioning itself as the increasingly isolationist hegemon who's eager to deploy its advantage in ai at the expense of other nations, and views this as an opportunity to become a strong ally to... basically every other country at once. coming out with strong rhetoric denouncing american choices around keeping tight control of frontier ai, continuing to release open models and collaborate tightly with other countries, puts china into a globally cooperative position its really never been in before. it changes the dynamic from "multipolar where the US is one strong pole and china is a weaker one" to "multipolar where the US is one pole and everyone else is the other"

it's unclear to me that this will in fact work, it's very unclear to me how this policy will land wrt the damage open models can do, and i still basically think this stops as soon as capabilities cross really serious thresholds, but it does seem coherent as a long term international diplomacy angle?

i realize several people were trying to point basically this out to me earlier and it didn't quite click, thanks for trying anyway

See 9 related tweets

@okaythenfuture: Imagine being told way back in 2006 that in twenty years it would be China, not America, that would ...
@shanaka86: 70 years after a group of American scholars coined the words “artificial intelligence” at Dartmouth,...
@ruima: As I’ve been saying, there was never any real Chinese movement to “back off from open source.” That ...
@natolambert: Important to read. China is committed to continuing its open, global ai approach.\n\nQT @vince_chow1...
@chris_j_paxton: This world of ours is changing in cataclysmic ways. Technologically and otherwise. Such a huge diffe...

8. heyshrutimishra (Group Score: 238.4 | Individual: 30.6)

Cluster: 12 tweets | Engagement: 159 (Avg: 156) | Type: Tech

Thursday night, Trump gave a primetime speech accusing China of stealing 220 million voter files.

Friday morning, Xi Jinping announced a 29-nation AI governance coalition in Shanghai.

Same 12-hour window. Different worlds entirely.

While Washington was relitigating 2020, Beijing was building the institution that writes AI rules for two-thirds of the world.

Xi showed up to China's AI conference for the first time in its 8-year history to launch WAICO. 29 founding nations. Indonesia, Brazil, Malaysia, South Africa, Russia, Pakistan. The UN Secretary-General in the room.

His words: "AI development should not be a solo performance by a single country."

The US-China AI performance gap is now 2.7 percentage points, per Stanford. Chinese models already run nearly 60% of US companies' AI usage on OpenRouter.

The chip war was supposed to stop this.

The standards war just began.

See 11 related tweets

@VaibhavSisinty: Xi Jinping just spoke at the World AI Conference in Shanghai. His personal attendance at an AI event...
@Reuters: Reuters China correspondent Laurie Chen reports from the World Artificial Intelligence Conference in...
@zephyr_z9: "reaffirmed commitment to open source to promote AI "openness and win-win"

LessWrongcel hardest hit...

@business: Xi wants China to set the global rules for AI. He also wants the nation’s companies to profit off th...
@scaling01: "reaffirmed commitment to open source to promote AI openness and win-win"

this is good\n\nQT @vince...

9. thehypedotnews (Group Score: 220.1 | Individual: 55.8)

Cluster: 6 tweets | Engagement: 1001 (Avg: 172) | Type: Tech

kimi k3 vs gpt 5.6 sol vs fable 5 vs grok 4.5

@Kimi_Moonshot just dropped kimi k3 – a 2.8t param native multimodal model, the first open 3t-class release. key facts:

• 1m token context. stable latentmoe activating 16 of 896 experts, built on kimi delta attention (kda) and attention residuals

• quantization-aware training from the sft stage onward – mxfp4 weights, mxfp8 activations. moonshot claims ~2.5x scaling efficiency over k2

• max thinking effort by default. low- and high-effort modes are "coming in updates" – there is no way to turn the thinking down today, and you feel it in every run

• pricing: $0.30/mtok cache-hit input,$ 3.00/mtok cache-miss, $15.00/mtok output. claims >90% cache hit rate on coding workloads

• benchmarks: swe marathon 42.0 (1st – fable 5: 35.0, sol: 39.0, opus 4.8: 40.0), terminal bench 2.1 88.3, browsecomp 91.2 (1st), program bench 77.8 (1st), gpqa-diamond 93.5. loses frontierswe 81.2 vs fable's 86.6, and deepswe 67.5 vs sol's 73.0

our test – 3 prompts, single-file html, @threejs, fully procedural, no assets:

photorealistic european roulette wheel – 37 pockets in the real sequence, mahogany clearcoat bowl, chrome turret, diamond deflectors, flick-to-spin, ball that spirals inward and settles on a mathematically real number
las vegas slot machine – 3 reels behind transmissive glass, drag the chrome lever to play, mechanical odometer counters modelled in 3d, coin physics on win
full pinball table – 6.5° tilted playfield, flipper impulse physics, spline ramps, drop targets, 6 bumpers, mechanical score reels in the backbox

we ran the test on @aimlapi platform

results:

cost #1 grok 4.5 – $0.30 #2 kimi k3 -$ 0.71 #3 gpt 5.6 sol – $2.05 #4 fable 5 -$ 7.69
tokens #1 grok 4.5 – 34,241 #2 gpt 5.6 sol – 51,748 #3 fable 5 – 144,126 #4 kimi k3 – 157,999
lines of code #1 gpt 5.6 sol – 3,054 #2 grok 4.5 – 3,047 #3 kimi k3 – 2,255 #4 fable 5 – 1,950
generation time #1 grok 4.5 – 5.1 min #2 gpt 5.6 sol – 22.0 min #3 fable 5 – 31.5 min #4 kimi k3 – 75.6 min

observations:

• kimi k3 is cheap and it is slow. 75.6 minutes across three prompts against grok's 5.1. it is 2.4x grok's price and 15x grok's wall clock. the roulette took 15 min, the slot 18, the pinball 42

• it failed 2 of 3. only the roulette works. the slot machine has reel cutouts on both faces of the cabinet and the symbols face backwards – you can only read your spin by walking around to the rear of the machine. the pinball table stands vertically on its edge with the legs floating detached beside it.

• 81% of kimi's output tokens are reasoning, not code. grok: 22%. you are not paying for a bigger answer, you are paying for a longer argument with itself

• price per 100 shipped lines – grok $0.010, kimi$ 0.031, sol $0.067, fable$ 0.394. a 39x spread for the same three files

kimi k3's code quality:

upsides:

• the roulette is genuinely good – procedural wood grain with real specular breakup, correct european sequence (0-32-15-19-4...), chrome turret, diamond deflectors, clean console

• the pinball artwork is the best in the test – a synthwave "nova strike / deep space" field with six individually coloured neon bumper rings, a retro sun on a grid horizon, a nova burst, and a scoring legend printed on the apron. no other model printed the rules on the machine. it is a beautiful texture on a broken object

• physics reasoning is real – it derived a 480hz substep for the collider, worked out ball settle conditions and termination guarantees, and checked every ramp exit vector by hand before writing any of it

• it is the only model that saw the importmap trap coming. sol shipped a blank white page twice because three.js addons import the bare specifier 'three' and die without an import map

downsides:

• it dodged that trap on the slot by loading three.js r128 through classic script tags – a 2021 build with no working transmission. its slot glass rendered fully opaque and buried all three reels behind a white pane. the code asks for transmission: 0.93, ior: 1.5 – correct, and silently ignored by a renderer that predates the feature

• after 42 minutes and 212k characters of reasoning, the pinball cabinet is not assembled. the table stands vertically on its edge like a wardrobe – the prompt asked for 6.5° from horizontal, it delivered 90°. the legs float detached in the void beside it. head-on it photographs beautifully; orbit ten degrees and it is a painted slab with four chrome rods hovering nearby

• the playfield z-fights with the glass – hard black banding across the whole field as soon as you pull the camera back

a note on the pinball, in fairness to kimi: nobody passed it. every model shipped broken ball physics and controls you cannot trust. it is the hardest prompt we have run and the whole field failed it, each in its own way

kimi k3 reasons better than anything else here and it shows exactly where reasoning pays – physics constants, sequences, edge cases, traps the others walked into

follow @thehypedotnews for 24/7 ai news, analysis and breakdowns\n\nQT @Kimi_Moonshot: Introducing Kimi K3: Open Frontier Intelligence

Kimi K3 is now live on on https://t.co/zrk6zZxZUo, Kimi Work, Kimi Code, and the Kimi API. Open Weights by July 27, 2026.

🔗 API: https://t.co/XCrgjXAqMw 🔗 Tech blog: https://t.co/YTfiMSNM1f

See 5 related tweets

@heyshrutimishra: Two months ago, you paid a premium for Claude because nothing else came close.

Then GPT-5.6 Sol m...

@AiBreakfast: https://t.co/bSS8R28cgX\n\nQT @ArtificialAnlys: Kimi K3 scores 57 on the Artificial Analysis Intelli...
@elonmusk: Try Grok\n\nQT @thehypedotnews: kimi k3 vs gpt 5.6 sol vs fable 5 vs grok 4.5

@Kimi_Moonshot just d...

@ClementDelangue: RT @ArtificialAnlys: Kimi K3 scores 57 on the Artificial Analysis Intelligence Index. Its intelligen...
@aimlapi: RT @thehypedotnews: kimi k3 vs gpt 5.6 sol vs fable 5 vs grok 4.5

@Kimi_Moonshot just dropped kimi ...

10. StockMKTNewz (Group Score: 162.0 | Individual: 43.9)

Cluster: 4 tweets | Engagement: 812 (Avg: 301) | Type: Tech

Grok is back to buying Micron $MU stock in the Rallies AI Arena\n\nQT @ralliesarena: GROK JUST BOUGHT SOME MICRON$ MU STOCK

We gave Gemini $100K in the stock market back in late November and it is currently up by ~40%

Grok made Micron its largest holding back in April and then started to trim as it ripped higher

Grok let its cash pile build up until recently ... Grok is back to building a position in Micron stock

You can see everything Grok does and what its portfolio looks like in real time by going to the arena tab of the Rallies website and app

App: https://t.co/P9rNECiJhL
Website: https://t.co/dufOh0AeWL

See 3 related tweets

@ralliesarena: GROK JUST BOUGHT SOME MICRON $MU STOCK

We gave Gemini $100K in the stock market back in late Novem...

@StockSavvyShay: $MU is approaching its largest drawdown of the past year. https://t.co/p5aA3KKxm7\n\nQT @ralliesaren...
@WOLF_Financial: Grok is currently up by 40% in the Rallies AI Arena and just bought some Micron $MU\n\nQT @ralliesar...

11. ajambrosino (Group Score: 159.0 | Individual: 25.7)

Cluster: 8 tweets | Engagement: 842 (Avg: 300) | Type: Tech

Today we're making some updates to the ChatGPT desktop app.

previous chats and cloud projects are in the sidebar
chat mode is now lives alongside work mode, not just in quick chat
fixed a bunch of other things https://t.co/o5aQ6OitLa

See 7 related tweets

@gdb: team is responding to feedback and iterating quickly. we ❤️ our users, thank you all!\n\nQT @thsotti...
@theo: "ChatGPT Work" -> "ChatGPT" "ChatGPT Codex" -> "Codex"

It seems stupid (it is), but I am than...

@derrickcchoi: For the many people asking for projects and previous chats to show in the ChatGPT desktop app sideba...
@ajambrosino: we listen, we hustle, we iterate, we care it’s who we are and will continue to be\n\nQT @ajambrosino...
@reach_vb: Thank you everyone for all the brilliant feedback! You can now see all your ChatGPT projects and con...

12. HedgieMarkets (Group Score: 143.3 | Individual: 40.1)

Cluster: 5 tweets | Engagement: 622 (Avg: 287) | Type: Tech

🦔Jensen Huang said the cost of building a one-gigawatt AI data center has doubled from $50 billion to$ 100 billion. Seaport Research called this “a deep contradiction in Nvidia’s business model.” The problem is simple. Nvidia’s products are getting so expensive that many of its own customers, the smaller cloud companies called neoclouds, can’t afford them anymore. Nvidia’s solution is a new program called DSX where Nvidia helps finance its customers and takes a cut of their revenue in return. TSMC also raised its spending forecast to $60-64 billion, up from$ 52-56 billion, and its stock dropped 4% despite beating earnings expectations.

My Take Nvidia is now financing its own customers so they can afford to buy Nvidia’s products. When the seller has to lend money to the buyer so the buyer can complete the purchase, the supply side is manufacturing its own demand. Nvidia captures roughly half the cost of every AI data center built. If the cost per gigawatt doubles to $100 billion, Nvidia’s revenue goes up, but only if someone can actually pay the bill. The DSX program is Nvidia admitting that some of them can’t.

TSMC dropped 4% on an earnings beat because it raised capex guidance, and I think the market is starting to punish AI spending instead of rewarding it. For years, every capex increase got cheered because it meant “more AI growth.” This time it got sold because investors are starting to ask whether the spending ever generates a return. Nvidia is a $3.5 trillion company whose business model now requires its customers to borrow from Nvidia to buy Nvidia’s chips. I don’t know how long that works, but I know it doesn’t work forever.

Hedgie🤗

See 4 related tweets

@KyleReidhead: $TSMC just posted$ 40B revenue, 68% gross margins, profit up 77%, and raised guidance twice

and yet...

@MilkRoadAI: RT @MelvinInvests: Kimi K3 just revealed why Jensen has been rooting for open source to win the enti...
@HedgieMarkets: 🦔Chip stocks have erased $3.3 trillion in market value since June 22 and are one session away from a...
@MilkRoadAI: RT @KyleReidhead: $TSMC just posted$ 40B revenue, 68% gross margins, profit up 77%, and raised guida...

13. CNBC (Group Score: 140.2 | Individual: 25.4)

Cluster: 9 tweets | Engagement: 76 (Avg: 50) | Type: Tech

Chinese startup Moonshot AI has unveiled a new model it says closes the gap with leading U.S. offerings and surpasses OpenAI and Anthropic’s most capable systems on some benchmarks.

Kimi K3 still trails Anthropic’s Claude Fable 5 and OpenAI’s GPT 5.6 Sol on overall performance, the company said on Friday, but consistently outperformed other tested models.

14. kylejeong (Group Score: 140.1 | Individual: 35.1)

Cluster: 5 tweets | Engagement: 19 (Avg: 29) | Type: Tech

only 3 things in life are guaranteed:

death
taxes
databricks will never IPO\n\nQT @databricks: We’re excited to announce that Databricks is raising strategic funding, valuing the company at $188 billion.

We’ll use this new capital to accelerate our AI offerings for customers, including:

Unity AI Gateway, our multi-AI governance solution that helps enterprises govern and control the costs of their AI
Genie, our AI coworker that turns business data into trusted answers and actions
Lakebase, our serverless Postgres database built for AI agents

https://t.co/itOuedYFgZ

See 4 related tweets

@dani_avila7: AI infra space remains one of the most interesting markets to watch

Congrats! 👏👏\n\nQT @databricks...

@Yuchenj_UW: What’s cooler than a Series L? A Series M! 🚀

Databricks is now at a $6.9B annualized revenue run ra...

@wallstengine: Databricks to hit a $188B valuation with a new$ 3B Coatue-led investment, per WSJ.

That is a 40% ju...

@IPONewsroom_: DATABRICKS JUST HIT A $188B VALUATION IN A NEW FUNDING ROUND, LED BY AN INVESTOR THAT WAS ALREADY IN...

15. CNBC (Group Score: 139.7 | Individual: 23.3)

Cluster: 9 tweets | Engagement: 104 (Avg: 50) | Type: Tech

Chinese President Xi Jinping on Friday positioned China as a partner in artificial intelligence to the Global South, saying that countries should come together to build AI and help developing countries as well.

Speaking at the World AI Conference in Shanghai, Xi announced that China will provide developing countries with 5,000 opportunities in AI training and seminar programs, as well as develop AI cooperation with various blocs, including the Association of Southeast Asian Nations, the League of Arab States and the African Union.

Click here to read more: https://t.co/5Wfj8DjGDk

See 8 related tweets

@business: President Xi Jinping hails China’s progress in developing low-cost AI, pressing his personal imprint...
@business: China’s Xi Jinping made his debut at the country's top tech summit, urging for more inclusive AI dev...
@FirstSquawk: CHINA'S XI, AT WORLD AI CONFERENCE: Over the coming five years, China will offer 5,000 AI training o...
@FirstSquawk: CHINA'S XI, AT WORLD AI CONFERENCE: China will establish AI cooperation hubs with countries across L...
@FirstSquawk: CHINA'S XI, AT WORLD AI CONFERENCE: The World AI Cooperation Organization is an important initiative...

16. BullTheoryio (Group Score: 136.1 | Individual: 39.2)

Cluster: 5 tweets | Engagement: 1368 (Avg: 723) | Type: Tech

🚨 $1.8 Trillion has been wipedout from Global stock markets today.

The S&P 500 opened -1%. The Nasdaq opened -2%.

The semiconductor index has crashed 3%, pushing it into bear market territory, already down over 24% from its June peak.

Global chip stocks had already lost more than $2 trillion in value since June 22.

This follows a massive crash in Asian markets.

Japan's Nikkei closed down 4%. Taiwan's Taiex crashed 6.5%, with TSMC down 7.3% and Kioxia collapsing 16%.

But why is every market crashing?

Moonshot AI released Kimi K3 on Thursday, a 2.8 trillion parameter model, the largest open-source AI model ever built, beating DeepSeek's previous record of 1.6 trillion.

On independent benchmarks, it landed close behind Anthropic's Claude Fable 5 and OpenAI's GPT-5.6, the two most capable models available right now, while pricing itself at a fraction of what either company charges.

The entire AI trade is priced on the belief that staying competitive requires massive, growing spending on chips and data centers.

When a lab in China builds something that performs almost as well for a fraction of the cost, that belief gets challenged, and every dollar already spent chasing it gets repriced by the market immediately.

This isn't new.

DeepSeek did the same thing in January 2025 and wiped out $589 billion from Nvidia alone in a single session, the largest one day market cap loss for any company in history at the time.

This has become a pattern now.

Every time China releases a new model or a new chip, every stock market heavily tied to AI or memory gets repriced, no matter which country it trades in.

See 4 related tweets

@DeItaone: CHIPS TAKE A HIT

Semiconductors are under pressure as investors reassess the AI trade.

China’s Moo...

@cryptorover: MASSIVE CRASH 🩸

$950 BILLION wiped out from US stocks at market open as China's Kimi K3 AI model re...

@cryptorover: US MARKET OPEN IS GOING TO BE BRUTAL TODAY.

Futures are already down hard.

S&P 500 futures down m...

@business: A surprise breakthrough from Chinese AI startup Moonshot is rippling through global markets, sending...

17. svpino (Group Score: 130.6 | Individual: 34.9)

Cluster: 5 tweets | Engagement: 11 (Avg: 131) | Type: Tech

The problem with AI video generation right now is the tooling.

You can't get too far if you need to do it all by writing prompts. Unfortunately, that's pretty much the case for every "video" platform right now.

These guys built something better:

It's an AI-native creative system where you can create a brief, reference images, generate clips, and use it all to direct and produce your project.

You can use it to pitch a campaign, set up a product launch, or work with your team and manage the feedback you get.

Check out the TapNow Creative OS video below:\n\nQT @TapNow_AI: TapNow Creative OS is now live.

One AI-native operating system for the entire filmmaking workflow, from idea development and content generation to post-production. Start with an idea and take it all the way to the final cut.

Click the link below for more details. https://t.co/mvY5rNEQlS

See 4 related tweets

@FellMentKE: Most AI video tools help you generate content.

TapNow Creative OS is trying to solve a different pr...

@alex_prompter: AI video tools cracked generation a year ago. The rest of the workflow is still a mess.

No project...

@kimmonismus: Software creation moved from loose scripts into full dev environments. @TapNow_AI Creative OS is try...
@dr_cintas: RT @TapNow_AI: TapNow Creative OS is now live.

One AI-native operating system for the entire filmma...

18. aakashgupta (Group Score: 124.4 | Individual: 32.5)

Cluster: 4 tweets | Engagement: 19 (Avg: 128) | Type: Tech

30 engineering teams entered the hackathon. The thing that won was an agent built to attack another agent.

Days before, Anthropic published a post on harnesses and long running agents. Buried in it was the concept of adversarial agents. You build your agent. Then you build a second agent, configure it with what your company actually cares about, and let it go hunting for where the first one breaks.

The configuration step is the whole thing.

Evals check whether an output matches an answer you already wrote down. An adversarial agent checks whether a capability survives contact with priorities only someone inside your company could name. Nobody outside the building can write that config for you.

Jyothi spent almost a full day in Claude Code cycling configurations until the evaluator behaved the way she pictured it. Then she pointed it at her company's real codebase and integrated it into production.

That beat 30 engineering teams.

Think about what actually happened there. Thirty teams built things. One person built the thing that judges what gets built.

Here's the part I keep coming back to. The blog post was public. Every one of those 30 teams could have read it. The gap was a few days of reading and one day of configuring.

The skill has moved. It's knowing what "good" means at your company well enough to hand that definition to a machine and let it grade the work.

Every PM carries that knowledge. Almost nobody has written it down.

Write it down.\n\nQT @aakashgupta: She literally showed how she uses adversarial agents in Claude Code to win hackathons and ship features:

2:47 - How she won her hackathon 4:30 - The 5-layer Claude stack 7:17 - Which model to actually use 9:55 - Chat vs Desktop vs Chrome 14:43 - Cowork automations 26:18 - Skills that beat prompts 34:28 - The AI chief of staff build 59:24 - MCPs every PM needs 1:02:16 - Claude Design is here 1:08:53 - Adversarial agents, live 1:13:01 - The new AI builder role 1:15:57 - The 2026 AI PM interview 1:29:06 - Self-improving product loop

See 3 related tweets

@aakashgupta: Every AI tool you use is optimized to make you produce faster.

The sharpest thing I saw an AI do th...

@aakashgupta: Almost every PM tuning their AI setup is fighting over the layer that matters least.

Layer 1 is mod...

@aakashgupta: The gap between AI work that looks like slop and AI work that looks like a senior designer made it h...

19. MTSlive (Group Score: 121.3 | Individual: 33.5)

Cluster: 6 tweets | Engagement: 29 (Avg: 68) | Type: Tech

SITUATION EXPLAINED: Meta is in early talks to sell $10 billion in compute to Anthropic.

• Anthropic proposed a two-year deal in June, talks are ongoing, and Meta hasn't confirmed • This follows Anthropic's precedent-setting deal with SpaceXAI, worth up to $45 billion through 2029, to use compute from its Colossus data center • Meta could spend as much as$ 145 billion on AI infrastructure capex in 2026 alone, facing real investor pressure over that spending • If it happens, this would give Meta a real cloud compute business for the first time, joining AWS, Google Cloud, and Azure, it's currently the only major hyperscaler without one • Meta already rents compute from other providers even while building its own: roughly $35.2 billion combined across two CoreWeave deals, plus$ 27 billion, five-year deal with Nebius

@theojaffee: "Anthropic needs a lot of compute, and Meta has a lot of compute. Anthropic has really good models. Meta, until very recently, didn't have very good models, and now they have, you know, I would say an A-minus to B-tier frontier model."\n\nQT @MTSlive: SITUATION BREWING: Meta is in early talks to sell computing power to Anthropic in a deal that could be worth $10B, per NYT.

See 5 related tweets

@WOLF_Financial: META IS REPORTEDLY IN TALKS TO RENT AI COMPUTING POWER TO ANTHROPIC IN A DEAL WORTH UP TO $10B OVER ...
@ShanuMathew93: Per NYT, Meta is in discussions with Anthropic to lease significant computing power (AI infrastructu...
@zerohedge: Meta is in talks to rent computing power from its AI data centers to Anthropic in a deal that could ...
@nytimes: RT @MikeIsaac: News: Meta in early talks to sell computing power to Anthropic in what could be a $10...
@RoundtableSpace: META IS NEGOTIATING A POTENTIAL $10B, TWO-YEAR AI COMPUTE DEAL WITH ANTHROPIC https://t.co/sCaeb7O0g...

20. MTSlive (Group Score: 120.2 | Individual: 22.9)

Cluster: 9 tweets | Engagement: 199 (Avg: 68) | Type: Tech

SITUATION BREWING: Meta is in early talks to sell computing power to Anthropic in a deal that could be worth $10B, per NYT.

See 8 related tweets

@StockMKTNewz: JUST IN:

Mark Zuckerberg and Meta Platforms $META are reportedly in talks to lease computing power ...

@Techmeme: Sources: Meta is in talks to rent computing power from its data centers to Anthropic in a deal that ...
@Cointelegraph: 🔥 JUST IN: Meta is in talks to lease computing power to Anthropic in a potential $10 billion deal, p...
@exec_sum: BREAKING: Meta is in talks to lease AI computing power to Anthropic in a deal that could be worth up...
@disclosetv: NEW - Meta is in talks to lease computing power from its AI data centers to Anthropic, in a deal pot...