Nvidia Ships AI Factories by Truck
Nvidia unveiled the DSX Pod AI Factory, packing seven Vera Rubin racks into a 28,000-kilogram shipping container with cabling and liquid cooling preassembled. Nvidia says it cuts the time for a new AI data center to start serving tokens from six months to 14 days. A single Vera Rubin rack can cost over $7 million.
OpenAI Models Land on Google Cloud
OpenAI is bringing its frontier models and Codex agentic workflows to Google Cloud’s Model Garden, letting enterprises use them where they already build. It extends OpenAI’s enterprise reach beyond Microsoft Azure and puts its models inside a rival hyperscaler that also sells Gemini, a notable shift in AI platform power.
Musk Says Optimus Production Has Begun
Elon Musk said Tesla has started early Optimus production, with most of a one-million-unit-per-year line installed at Fremont. He tempered expectations: the robot’s AI, electromechanical design, reliability and usage life still need work, which he expects to be solved by the end of next year. He still calls Optimus the biggest product on Earth.
Jensen Huang Backs California Billionaire Tax
Nvidia CEO Jensen Huang endorsed California’s proposed billionaire tax, even though it could cost him roughly $8 billion. “I don’t mind paying taxes. The more you make, the more taxes you pay,” he said, calling himself the ultimate American Dream. The comment lands as California weighs taxing its richest tech founders.
Waymo Heads to Madrid
Waymo is laying the groundwork for fully autonomous ride-hailing in Spain, starting with the Region of Madrid. The Alphabet unit has completed more than 20 million trips and serves hundreds of thousands of paid rides each week, and Madrid extends its push to export the Bay Area robotaxi playbook to European cities.
Sequoia Makes Blacksmith a Unicorn
Blacksmith, which speeds up continuous integration (CI) for software increasingly written by AI, raised a $66 million Series C led by Sequoia at a $1 billion valuation. That is up from $550 million two months ago and $60 million a year ago. Since January, ARR has grown 17x, while weekly CI jobs have surged from 5 million to more than 90 million.
Arda Launches With $232M for Factories
Arda Industries, which builds AI software that makes factory production lines more adaptable, launched with $232 million in funding. Accel, Altimeter, Founders Fund and Khosla Ventures led the round, joined by 8VC, SV Angel, Felicis and XYZ Venture Capital. The bet: AI that retools lines fast is the missing piece of American reindustrialization.
Intel Sets High-NA EUV for 2028
Intel has begun installing production tools at its Arizona Fab 62, which is expected to begin operations by the end of 2028. Separately, Intel has already shipped high-volume logic chips using High-NA EUV on select layers of its 18A Panther Lake processors, with broader adoption planned for future nodes. Its Ohio fab remains years behind the original schedule, and Intel has yet to announce a major external customer for its 14A process.
Memphis Rejects Data Center Moratorium
Memphis city councilors voted down a proposed 18-month moratorium on new data centers, opting instead for a 135-day pause on new projects. The city hosts xAI’s large AI compute cluster, making it a bellwether for how local governments balance AI infrastructure growth against power, water and community pushback.
Kevel Raises $52M to Buy Rivals
Retail media infrastructure company Kevel raised a $52 million Series D and plans to use part of it for acquisitions, according to CEO and founder James Avery. With many small retail media players competing for retailer budgets, Kevel is betting consolidation is coming and wants to be the buyer, not the target.
OpenClaw’s Creator Burns $10M a Month
Peter Steinberger, who created the open-source agent OpenClaw and now works at OpenAI, estimates he burned about $10 million in AI tokens last month. In May, he posted a monthly bill of $1.3 million. OpenAI pays. At a Cloudflare conference, the host noted that Steinberger has “a virtually unlimited token budget, which not all of us in the room have.” As the room murmured, Steinberger added: “not all of that is useful.”
OpenAI is showing what agents cost when nobody counts. Steinberger is the best-known builder of the personal agents every giant now sells, and his spending has grown about eightfold since May. Part of it, by his own account, buys nothing. His employer is racing to raise money before Anthropic goes public; at this pace, his habit costs it $120 million a year. Customers who copy his workflow will pay for the waste themselves.
At MyClaw, we run OpenClaw alongside rival Hermes, and the newest OpenClaw wins. That may be what $10 million bought.
Humans Were the Bug
PHP, the language behind most of the web, rushed out updates for two flaws, each rated 9.2 out of 10 for severity. Either can let an attacker take over a server without logging in, and nearly every version since 2012 is exposed. Versions older than 8.2 will get no fix. AI helped find the flaws: security firm Calif used OpenAI’s GPT-5.6 Sol, and Google-owned Wiz used Gemini. Governments in the Netherlands, Finland and France pulled sites until they could patch.
AI has made humans the bottleneck, and PHP’s security team is the proof. Its own 2024 audit found the root of one flaw. The team needed every member to agree, could not pick a fix and kept stalling without a working attack. AI built the attack. Attackers were hitting Amazon’s decoy servers four days before the patch. OpenAI expects such findings to “increase sharply,” all landing on the same few engineers.
PHP’s remedy: majority votes on its security team. It took a global outage to retire unanimity.
Giants Wants Free Chatters to Pay
Anthropic opened a waitlist for Claude Duet, a personal assistant you simply chat with: ask for something and it gets it done. That makes five giants in the same race. Google’s Spark, SpaceXAI’s Grok Bot, Meta’s Muse and OpenAI’s Dots all arrived within five months. Meta’s Muse is free and shot to the top of the App Store. The rest sit behind paid plans; OpenAI’s Dots starts at $100 a month.
The giants already have a billion people chatting with AI for free. Now they want those people paying for AI that does the work. Agents are where the money is, and almost nobody uses them: roughly one person for every hundred who chat. The personal assistant hides the agent inside a familiar chat window. Muse shows the catch. By one estimate, four in five users drop it within two weeks.
The giants are betting a billion people will pay to stop doing chores themselves. So far, most would rather keep chatting for free.
Zuckerberg Picks an Old Fight
Meta will hand every shareholder a non-voting share for each one they own. The reason is the AI bill. Meta expects to spend up to $140 billion this year, much of it on AI data centers, and its free cash flow has turned negative. The stock lets it raise money without diluting Mark Zuckerberg’s control. His own holding companies approved it in writing; no other shareholder votes. Shareholders sued over a similar plan in 2016, and he dropped it.
The AI race now writes Meta’s share structure. Buybacks, which kept Zuckerberg’s stake from shrinking, stopped as cash went to AI. Bond demand cooled too, on a July sale for a Texas data center. That leaves stock. The new class lets Meta pay engineers and buy companies with it while one man keeps the votes. The filing even cites keeping him near Meta Superintelligence Labs.
When he dropped it, he said he could keep control “for 20 years or more” without new shares. The AI bill cut that to nine.
DeepSeek open-sources its production inference engine to cut compute costs
DeepSeek just open-sourced DeepInfer, the actual engine running their models in production. Not a demo. Not a research paper. The real thing.
Think of an inference engine as the software layer that takes a trained AI model and makes it answer questions fast, at scale, without melting your server budget. That is what DeepInfer does.
Here is what is inside:
- Speculative decoding via DSpark: a draft model guesses tokens ahead, a confidence filter trims ~30% of checks, netting a 4-7% throughput gain
- Distributed inference to split work across multiple chips
- Structured outputs so responses follow a set format
- Runs on Nvidia SM100 and Huawei Ascend 950 chips
They also dropped deepseek-recipe, a token encoding library that mirrors the official DeepSeek API exactly. Self-host it and your outputs behave identically to the hosted version. No more subtle mismatches.
Both repos are on GitHub under Apache 2.0. Future model code lands here too.
Anthropic previews Claude Duet, a proactive agent that works while you’re away
Most AI tools sit there waiting for you to type something. Claude Duet flips that. It comes to you.
Anthropic just launched a research preview of Duet, a standalone app where Claude acts as a proactive agent. Think of it less like a chatbot and more like a background worker that monitors your apps, surfaces what matters, and handles tasks while you are away.
Here is what it actually does:
- Sends you a morning brief from your connected apps
- Monitors email, calendars, and Slack, then pings you only when something needs you
- Connects to 6,000+ tools including invoicing and scheduling
- Keeps running in the cloud even when your device is off
- Asks your approval before sending messages, payments, or deleting anything
It runs on Claude Opus 5.5 and pulls in your existing Claude memory and connected apps, so setup is minimal. It never sees your passwords, and a dedicated safety layer guards against leaking sensitive info.
Access is via waitlist for Pro and Max subscribers in the US, at no extra cost, though background tasks count toward your usage limits.
Unitree unveils a kung fu humanoid robot with 8-hour battery life
Unitree just dropped the H2-X Kung Fu Legend Guardian, a humanoid robot built for security and rescue. Think less warehouse helper, more bodyguard with a black belt.
Here is what it can actually do:
- Run for 8 hours on a single charge, nearly triple the base H2’s 3-hour battery life
- Jump off 4-meter platforms and land without falling over
- Lift cars to free people trapped underneath
- Execute real martial arts moves using an AI model that predicts opponent actions in real-time, no human controlling it
- Pack 2,070 TOPS of onboard AI compute, roughly the processing muscle needed to run complex decisions locally
It stands 182cm tall, weighs 70kg, and has 31 movable joints. Unitree actually warns you to stay 2-3 meters away while it operates. That tells you everything about how much force those joints generate.
No pricing or release date announced yet. The base H2 starts at $29,900 for context.