有个判断，未来如果 AI 想要渗透或者让 C 端用户能够付费的话，只有一个方向就是游戏。我没见除了衣食住行外，能够让全民付费的情况除了游戏还有啥？

Bitcoin volatility has collapsed, with 90d realised volatility at the 8th percentile of its trailing four-year history. The 180d close range was at the 12th percentile, while DVOL is very low as well. Bitcoin has entered an extreme low-vol and transition state. These tend to resolve over the next 30-60days. What matters more is how the range ultimately resolves. The overwhelming majority of upside breaks have been sustained with strong forward median returns. Downside breaks have been mixed, with several violent reversals. Which way this resolves largely depends on the outlook for liquidity.

SpaceXAI's Grok 4.6 scores 61 on the Artificial Analysis Intelligence Index, joining the frontier in line with GPT-5.6 Sol, with standout agentic performance at lower cost Grok 4.6 gains 5 points over Grok 4.5 on the Intelligence Index just over one month after its release, or +23 points compared to Grok 4.3. This brings SpaceXAI back to the intelligence frontier alongside OpenAI, behind only Anthropic. Key takeaways: ➤ Grok 4.6 joins the frontier of the Artificial Analysis Intelligence Index: It scores 61, in line with GPT-5.6 Sol (max), behind Claude Opus 5 (max, 63) and Claude Fable 5 (max with fallback, 62), and just ahead of Kimi K3 ➤ Strong agentic performance: Grok 4.6 achieves a GDPval-AA v2 Elo of 1753, behind only Claude Opus 5 and with overlapping confidence intervals with Claude Fable 5 and Qwen3.8 Max. It scores 50.7% on 𝜏³-Banking, among the top two scores alongside Qwen3.8 Max (51.3%), and 88.4% on Terminal-Bench v2.1, in line with the leading models ➤ Frontier-level intelligence at lower cost: Headline pricing is unchanged from Grok 4.5 at $2/$6 per 1M input/output tokens, 60%+ below Claude Opus 5 ($5/$25) and GPT-5.6 Sol ($5/$30). It cost $0.84 per task, the same as Kimi K3 with slightly higher intelligence, placing it on the Intelligence vs. Cost per Task Pareto frontier ➤ Grok 4.6 sits at Fable 5-tier on AA-Briefcase, our private benchmark of long-horizon agentic knowledge work tasks, with an Elo of 1577 - behind the Claude Opus 5 family. It is notably turn-efficient, completing tasks in ~53 turns and ~0.5B input tokens on average vs. ~103 turns and ~2.0B input tokens for Claude Opus 5 (max) Other model details: ➤ Context window of 500k tokens (unchanged from Grok 4.5) ➤ Pricing of $2/$6 per 1M tokens of input/output; cache hits discounted to $0.5 per 1M tokens, an increase over Grok 4.5’s $0.3 per 1M tokens for cache hits Congratulations to @spacexai and @elonmusk on the release!

Traditional server architectures fix memory capacity at deployment time. As AI models grow larger and workloads become more memory-intensive, that rigidity creates stranded resources, inefficient utilization, and a well-documented memory wall bottleneck. The Marvell Structera S CXL switch addresses this by enabling memory pooling and composable infrastructure across servers. The industry's first CXL 2.0 switch, the Structera S 20256 delivers 256 lanes, up to 2 TB/s switching capacity, and ultra-low latency data paths optimized for CXL memory traffic, decoupling memory from compute so resources can be dynamically allocated where workloads require them. See the product brief here: mrvl.co/45DT8jj

We’ve advanced AMIE, our research medical AI system, to support real-time video consultations. Our latest study demonstrates expert-level performance in a randomized, controlled trial with 300 simulated consultations. Read the blog to learn more: goo.gle/4qa41m3 youtu.be/wa6xwvk8rPA

$RKLB just posted record $2.36B of backlog and more than $1B of new launch and Space Systems contracts showing the business underneath Neutron is already scaling across rockets, satellites and components. CEO Peter Beck says “the window for an end year launch is narrowing” because Rocket Lab would rather qualify Neutron for reusability and high cadence than rush Flight 1 since the real moat comes from making medium-lift launch repeatable instead of simply proving it can fly once. If Rocket Lab gets that right then Neutron becomes the missing piece in a full-stack space platform that can design satellites, manufacture key components, build spacecraft, launch and operate constellations and eventually layer on recurring communications revenue through Iridium.

China’s humanoid robot makers commanded more than 97% of global shipments in the first half of 2026, according to new industry data affirming the country’s early lead against US rivals in the burgeoning field. bloomberg.com/news/articles/…

FABLE 5 CAME BACK NERFED. We re-ran the July 1st version of Claude Fable 5 on BridgeBench. The results are brutal: Debugging: 86.2 → 25.9 Refactoring: 73.6 → 38.4 Hallucination: 75.9 → 61.7 The new guardrails are kicking in on way too many tasks and falling back to Opus 4.8. This is not the model that got banned. Anthropic owes everyone an explanation.

关于 Meta 云业务这件事，再给大家提供一个角度。 云业务具体来看有两种，一个是类似于多数 Neocloud 的出租裸金属算力，另一个是类似于 AWS、Azure 这种全栈云服务。 如果 Meta 云做的是前者，那么对整个 AI 大科技 Capex 以及上游半导体订单，并不会构成系统性的利空。 这和前段时间 SpaceX 把 Colossus 1 算力租给 Anthropic 非常像。出租算力并非行业需求差，如果 A 厂不缺算力，它不会去找 SpaceX，如果各家 AI 公司不缺算力，Meta 也没必要搞云业务。 如果算力真的已经全行业过剩，GPU 云价格应该会崩溃下跌，而不是越来越多公司在到处找容量。 而如果 Meta 云做的事后者，市场对于 Meta 放缓资本开支就更加无从谈起了。 做这种 AI 云基础设施平台，不仅不能大幅缩减资本开支，反而需要更多的基础设施投入，Capex 还会存在上行的动能。 从这个角度来看，Meta 出租算力应该是结构性错配，而非全行业过剩。 半导体恐慌下跌，本质上还是因为涨太多了，价格是主因，炒作利空只是催化。

Woah. Nvidia $NVDA just created a new line of business for themselves. So, all those neoclouds like $CRWV $NBIS $IREN $APLD $SPCX that have been getting deals with hyperscalers worth billions? It’s because demand for compute, according to Jensen, is growing at a level that is beyond imagination. So, companies need to secure more compute. But, many of these neoclouds are struggling to finance large GPU deployments, even after securing long-term compute demand. So…Nvidia is going to help them out and share in the upside. “This new model enables AI clouds to procure NVIDIA infrastructure for AI-native, enterprise and ISV customers through economic alignment with a revenue-sharing and credit-support model. Through the partnership, AI clouds will sell NVIDIA-powered cloud services, with NVIDIA earning both standard product revenue and a share of the cloud revenue on the supported capacity. This structure accelerates adoption of NVIDIA platforms among the high-growth, high-conviction AI native sector, and provides NVIDIA with a recurring, usage-linked earnings stream.” Looks like Nvidia is going to make sure the best neoclouds don’t fail and this also shifts from a one-time GPU sale to a recurring, usage-based revenue stream…which just creates many more longer-term monetization opportunities.

No sector in history has ever been this large: The Information Technology sector now accounts for a record 39% of the S&P 500's total market cap. This percentage has more than doubled since the 2020 pandemic. This figure is also now above the 2000 Dot-Com Bubble peak of ~33% and the ~31% peak reached by the Energy sector in the 1980s. Including internet retailers and digital media platforms such as Amazon, $AMZN, and Netflix, $NFLX, tech now accounts for a record 50% of the S&P 500's market value. By comparison, this same group accounted for just 29% of the index at the Dot-Com Bubble peak. Tech is all that matters.

CNBC: How GE Vernova builds the massive gas turbines powering the AI data center boom > GE Vernova’s $GEV order book is full until 2029, with orders going as far out as 2031. > Hyperscalers like Amazon, Google, Microsoft and Oracle are lining up to buy the company’s gas turbines to help power AI data centers. > Executives from nearly every major hyperscaler have walked the floor of the factory, according to a person familiar with the visits, who asked not to be named because the details are not public. > The turbines are massive, at 31 feet tall and weighing 280 tons. One turbine can power roughly half a million homes. > Microsoft just bought seven of them to power its data center in Texas. At 2.7 gigawatts, it’s enough electricity to power about 3 million homes. > "About 20% of our gas power order book is going to a data center, artificial intelligence-type of application." - Koziner, Chief Commercial and Operations Officer of GE Vernova. > One turbine can cost more than $250 million, according to industry estimates. The price has soared, up 300% in the last 3 years, according to analysts at Melius.

Another point on why $AMZN is a no-brainer long here: 
It's going to have the largest self-built data center capacity in 2030 per WSJ. 
$GOOG will be the closest in overall capacity, but a quarter of its capacity will be leased, meaning it'll be paying a marked-up price. 
$AMZN will have a substantial cost advantage in its inference business as the majority of its capacity will be owned, so it'll be paying no markups to neoclouds. Its custom silicon will further deepen this cost advantage. No brainer long at all time low valuations.

HEAR ME OUT. Robotics will be one of the biggest investment themes of the next decade. Everybody knows the trend. But most investors are still looking at the wrong stocks. Here are my Top 5 robotics plays: 1. $AEVA A 4D LiDAR sensing technology developer for automotive, robotics and broader Physical AI applications. Revenue growth is exploding, with multiple future quarters expected above +90% YoY. Among other things $AEVA is backed by NVIDIA, Daimler Trucks or Airbus.