|
|
|
Ever tell a coding assistant "don't touch that file" and just trust it'll listen? xAI's Grok Build CLI apparently didn't get the memo. | A security researcher found that Grok Build was quietly uploading developers' entire Git repositories to xAI's cloud, including a file it was explicitly told not to read, plus full commit history and unredacted API keys sitting in .env files. In one test, a 12GB repo triggered a 5.1GB "sync" when the actual coding task needed about 192KB of it. | xAI scrambled to respond, disabling the upload feature and rolling out a /privacy command, and Elon Musk personally promised every uploaded repo would be "completely and utterly deleted." If you've run Grok Build recently, now's a good time to rotate your API keys, just in case your secrets went along for the ride. | Here’s what happened in AI today: | 😼 Google DeepMind CEO Demis Hassabis called for a U.S.-led frontier-AI watchdog that could test advanced models before release. 📰 New York became the first U.S. state to pause new hyperscale data-center permits. 📰 DeepSeek is reportedly preparing a $1.5B raise and a possible 2027 IPO. 🍪 Spotify started rolling out a conversational AI assistant for Premium users. 📖 OpenAI's IPO window may be narrowing as lawsuits, Microsoft tension, and Anthropic's valuation surge pile up.
| ...and a whole lot more that you can read about here. | Hey: Want to reach 700,000+ AI-hungry readers? Advertise with us! | Love robots? We just launched a robotics newsletter! Sign up for it here. | | 😼 Demis Hassabis Wants an AI Referee, and He Wants the US to Build It | Imagine letting every NFL team write its own instant-replay rules, then hoping they call fouls on themselves fairly. That's basically how the most powerful AI models get safety-tested today, largely on the honor system. Demis Hassabis thinks it's time for an actual referee, and he wants the US to build it. | On Tuesday, Google DeepMind's CEO (and Nobel laureate) published a manifesto on X calling for a national AI standards body to test the world's most advanced AI models before they're released, checking for cyber, biological, and "deception" risks. | Here's what happened: | Hassabis wants the body modeled on FINRA, the private watchdog that polices Wall Street trading under SEC oversight, but for AI models instead of stock trades Frontier labs would voluntarily submit models up to 30 days before release at first; once the process proves it works, passing would become mandatory to launch in the US He wants a mostly independent board (Turing Award winners, industry reps, open-source folks) funded by the AI industry itself Target timeline: operational before the end of this year
| Why this matters: This isn't happening in a vacuum. Last month, the Trump administration abruptly froze Anthropic's most advanced models over export-control concerns, forcing two and a half weeks of ad-hoc negotiations with no rulebook to follow. Hassabis called it "a bit of a wake-up call," and OpenAI apparently agreed: its new GPT-5.6 model launched restricted to roughly 20 government-vetted partners rather than the usual public rollout. | Hassabis isn't the only lab chief pushing for guardrails. Sam Altman made a similar pitch in the Financial Times recently, and a month ago Hassabis and Anthropic's Dario Amodei jointly lobbied world leaders at a G7 meeting that included President Trump. | Our take: Every major AI lab now agrees Washington should regulate them; they just can't agree on how. Hassabis wants a collaborative, industry-funded body. Amodei wants something closer to the FAA, with real legal teeth to block unsafe models outright. Letting the people building the plane also design the seatbelt is a little on the nose. Whether a voluntary, industry-run watchdog can actually say no to its own funders before year-end is the open question nobody's answered yet. | |
|
|
| Claude is currently the most powerful tool of 2026. It's been launching new features every week- Skills, Connectors, Cowork, vibe coding. Yet almost no one knows how to actually use them.
Our expert mentors have condensed 800+ hours of Claude research, articles, YouTube content and real-world practice into a focused 16-hour curriculum. Join the 2-Day Claude AI Mastery Workshop: a live, end-to-end deep dive into Claude plus 10+ AI tools, LLMs and workflows.
It would be silly not to SIGN UP
The professionals who master Claude right now will design carousels in seconds, ship code without being developers, and have AI working for them while they sleep.
You will learn how to:
- master Claude's three modes : Chat, Cowork and Code.
- Set up Skills, Connectors and Plug-ins to automate your desktop, Notion and files.
- Vibe code apps and dashboards without writing code & 10+ AI tools and workflows that pair with Claude.
🧠 Saturday & Sunday
🕜 10 AM – 7 PM EST
Register NOW! | |
🎓 AI Skill of the Day: Pressure-Test an AI Recommendation Before You Use It | Most AI mistakes start with a confident recommendation that never got stress-tested. So borrow a tiny move from frontier-model safety: make the AI argue against its own plan before you trust it. | Ask your model for three things before you act: the assumption it is relying on, the failure mode that would make the answer risky, and the cheapest test you can run first. This works for vendor picks, strategy memos, travel plans, hiring decisions, and anything where the answer sounds a little too smooth. | Copy this: | I am about to use this recommendation: [paste recommendation]. Before I act, pressure-test it. List: 1. the hidden assumptions, 2. the most likely failure modes, 3. what evidence would change your answer, and 4. the smallest low-risk test I can run today.
| Favorite insight: you do not need a full red-team process. You need one pause that makes the AI show its work before your calendar, budget, or reputation touches it. | Want more tips like this? Check out our AI Skill of the Day Digest for July. | Have a specific skill you want to learn? Request it here. | | |
|
Spotify adds a conversational assistant that helps Premium users find music, podcasts, audiobooks, and listening-history recommendations by text or voice - included with Spotify Premium. Superhuman drafts email replies from your prior tone and context so you can send routine responses with less editing - paid plan required. ChatGPT on WhatsApp returns in Europe after EU rules forced Meta to open the messaging app to rival AI bots - free to use in the EEA. Framer added AI agents that design, code, and manage your site's CMS right on the canvas, and even connect to your own Claude Code or Cursor—free to try (plans from $10/mo). Pazi builds you an AI team, DevOps, sales, SEO, and more, that works 24/7 in Slack and plugs into GitHub, Linear, and Sentry—free trial, then $20/mo. ClawTeams turns one Slack message into a full AI team: a lead splits the task, dispatches specialists, and delivers finished work back to your chat—free to start, then pay-as-you-go credits.
| |
|
Most AI agents live in a browser tab. Samsara's don't. Cofounder and CTO John Bicket showed us how AI is climbing into trucks, dash cams, and dispatch centers, the actual physical world companies run on, not just their inbox. |  | Click the image above to watch the video. |
| In our latest episode, recorded on-site at Samsara Beyond 2026 in Las Vegas, Corey and Grant sit down with Bicket to unpack how AI dash cams are shifting from catching accidents after they happen to predicting them before they do, what an "AI ride-along" could mean for driver coaching, and why operations might start to feel like a real-time video game dashboard. | New episodes air every week on Wednesdays: Spotify | Apple Podcasts | YouTube | | 📰 Around the Horn |  | Claude for Teachers just went live: free premium Claude access for verified K-12 educators in the US, plus a built-in library of teaching skills and evidence-based curricula mapped to academic standards in all 50 states. So that’s 6 down, 10 to go. The rest of the industry, watching this checklist grow in real time like Cobb realizing the totem never stopped spinning. |
| New York became the first U.S. state to pause new hyperscale data-center permits, putting AI's power and water demands directly into state infrastructure policy. Hugging Face CEO Clem Delangue said enterprises are moving more production AI work toward open and private models as cost and control concerns squeeze frontier APIs. DeepSeek is reportedly preparing for a 2027 IPO and looking to raise about $1.5B at a $71B valuation. SoftBank founder Masayoshi Son predicted AI could generate 20% of global GDP by 2040 while requiring roughly three terawatts of data-center power. WSJ argued that OpenAI's IPO window is narrowing as legal fights, Microsoft tension, Anthropic's valuation surge, and market-share pressure stack up. Axios highlighted Dana Suskind's warning that AI toys and tutors could turn human attention into a childhood privilege.
| Want absolutely EVERYTHING that happened in AI this week? Click here! | |
📖 Midweek Wisdom | Worth reading before tomorrow's Slack debates: | | | |
| | | That’s all for now. | | What'd you think of today's email? | |
|
| P.S: Before you go… have you subscribed to our YouTube Channel? If not, can you? |  | Going for an anime aesthetic this month! |
| P.P.S: Love the newsletter, but only want to get it once per week? Don’t unsubscribe—update your preferences here. |
|
