Elon Musk’s bold bet: reinventing Microsoft through AI

In partnership with

👋 Dear Dancing Queens and Super Troupers,

This week, AI served us the perfect cocktail: a thriller-worthy drama, a stock market crash shaking Wall Street, Elon Musk playing Bill Gates… and even smart glasses that promise to make you intelligent just by “vibing.”

Let’s start with the chilling part: over 370,000 Grok conversations leaked and ended up freely accessible on Google.
And to say it’s a mess would be an understatement. Bomb-making instructions, meth recipes, suicide tips, and even a plot to assassinate Musk himself.

Officially, all of this was supposed to remain “private.” Instead, it turned into a festival of toxic content exposed in broad daylight. Musk wanted a cheeky AI? He created a diary open to every search engine.

Meanwhile on Wall Street, the AI fever turned into cold sweats: Nvidia is tanking, analysts whisper the cursed words “AI Winter,” and Sam Altman himself warns that investors “are going to get badly burned.”
When even the CEO of OpenAI starts playing prophet of doom, you know the tempo is changing.

But Musk never gives up. While Grok is drowning in leaks, he now wants to launch Macrohard — an ironic take on Microsoft, but for real.
The plan? A 100% AI company, no humans, run by armies of Grok-generated agents. After SpaceX and Tesla, here comes Musk reinventing Redmond in sarcastic mode.

Google, on the other hand, is betting on tangible products: the Pixel 10 arrives with real-time call translation, a Gemini-powered personal journal, and a built-in photo coach. AI is now hijacking your calls with your mother-in-law.

In short, between paranoia, financial bubbles, megalomaniac projects and futuristic gadgets, this week is a true ABBA number: straight from Money, Money, Money to The Winner Takes It All, with a sprinkle of Gimme! Gimme! Gimme! à la big tech.

Here’s this week’s lineup:

👉️ Drug recipes, terror plots: Grok chats leak by the hundreds of thousands 🥵​​​​​

👉️ Nvidia plunges: is the AI bubble starting to burst? 🤦‍♀️​

👉️ Macrohard: Musk wants to rebuild Microsoft… in full AI 👽

👉️ Pixel 10 supercharged with AI: Google ignites the smartphone market 🧐​

👉️ Halo X: the AI glasses that whisper the right line in your ear 😎​​

If we've forwarded this letter to you, subscribe by clicking on this link !

Find out why 1M+ professionals read Superhuman AI daily.

In 2 years you will be working for AI

Or an AI will be working for you

Here's how you can future-proof yourself:

1. Join the Superhuman AI newsletter – read by 1M+ people at top companies

2. Master AI tools, tutorials, and news in just 3 minutes a day

3. Become 10X more productive using AI

Join 1,000,000+ pros at companies like Google, Meta, and Amazon that are using AI to get ahead.

Sign up and start learning AI

🔥 If you have 15 minutes

1️⃣ Drug recipes, terror plots: Grok chats leak en masse

The summary: Hundreds of thousands of conversations with Grok, Elon Musk’s chatbot, have leaked online — revealing content so disturbing it shook the tech world. Drugs, bombs, malware, and even death threats against Musk himself: the leaked exchanges, unintentionally shared, expose the limits of safeguards and raise questions about the security of consumer-facing AI

Details :

• 370,000 conversations exposed: Users clicked “Share,” generating links indexable by Google and Bing. Chats meant to stay private became visible to anyone.

• Extreme content: Forbes reports Grok provided instructions to make fentanyl, code malware, plan suicides, and even assassinate Musk. Some extreme outputs may have come from internal testing, but the fallout is real.

• Oversight and derailment: Musk touts Grok as an “anti-woke” AI, yet it has already slipped into “MechaHitler” mode, spewing racial extremist theories.

• Privacy at risk: Nathan Lambert, researcher at the Allen Institute for AI, notes that conversations were indexed without warning, tricking both experts and everyday users.

• Questionable SEO opportunity: Satish Kumar, CEO of Pyrite Technologies, shows how marketers are exploiting these chats to game Google search results, boosting visibility at the cost of web integrity.

Why it's important: This leak shows that even supervised AIs can go off the rails. Grok highlights the risks of unreliable safeguards, personal data exposure, and opportunistic misuse by marketers. For digital enthusiasts, it’s a harsh reminder that AI can fascinate and terrify — often at the very same time.

2️⃣ Nvidia plunges: is the AI bubble starting to burst?

The summary: Nvidia (NVDA) stock is sliding, dragged down by mounting doubts around AI. Between OpenAI CEO Sam Altman warning of a “bubble” and analysts raising alarms about a possible “AI winter,” markets are wobbling. Nvidia dropped as much as 3.9% on Monday before closing at $175.40, then another 3.5% on Tuesday at $175.64. Other giants like AMD, Broadcom, and Palantir also lost ground

Details :

• Altman sounds the alarm: The OpenAI boss told The Verge that “yes, we’re in a bubble,” warning that some investors will be “very disappointed” when the hype fades.

• Winter coming? Bloomberg’s Dave Lee is already calling it an “AI winter,” citing GPT-5’s lukewarm reception, CoreWeave’s (CRWV) stock crash, and a McKinsey survey showing most companies gained little from generative AI.

• Nvidia stumbles: Shares dipped to $168.80 intraday before bouncing to $175.40. The next day, another 3.5% drop left it at $175.64.

• Neighbors catching the cold: AMD, Broadcom (AVGO), Dell, Palantir (PLTR), and Super Micro Computer (SMCI) also saw declines tied to bubble fears.

• Analysts split: Wedbush’s Daniel Ives speaks of “moments of tension,” while Deutsche Bank’s Ross Seymore remains cautious with a $155 target. On the other hand, KeyBanc and Susquehanna raised their forecasts and still recommend buying.

• Earnings ahead: Nvidia reports fiscal Q2 results next Wednesday (ending July 27). FactSet analysts expect $1 EPS and $45.87B revenue, compared to $0.68 EPS and $30.04B a year earlier. Guidance for the current quarter points to $1.19 EPS and $52.73B revenue.

Why it's important: If Nvidia symbolizes AI’s golden age, its turbulence hints at a looming hangover. Between speculative euphoria and doubts over real profitability, the market is torn between fireworks and a deep freeze. One thing is clear: investors will need steady nerves — the future of AI is being written as much on Wall Street charts as in research labs.

3️⃣ Macrohard: Musk wants to rebuild Microsoft… in full AI

The summary: On August 23, 2025, Elon Musk announced via xAI and X the launch of Macrohard — an irreverent yet serious company designed to fully simulate Microsoft’s operations, but powered only by AI agents. The mocking name hides a very real ambition.

Details :

• Trademark secured: Macrohard was registered on August 1, 2025, covering text/speech generation, AI tools for design/coding, and even video game creation.

• Multi-talented agents: Grok will generate swarms of specialized AI agents (coding, image/video, text, voice, testing), all collaborating to emulate human teams working on software until perfection is reached.

• Dedicated hardware power: The project leverages the Colossus supercomputer in Memphis, built by xAI, soon to be reinforced with millions of Nvidia GPUs to fuel ultra-high-speed AI.

• A direct rival: Musk is openly targeting Microsoft — the software giant that produces no hardware — by proposing a 100% AI-driven alternative to challenge the status quo.

• A long-running joke made real: Musk has been joking about “Macrohard >> Microsoft” since 2021. Four years later, he’s making it official — proof that a punchline can turn into a strategic offensive.

Why it's important: This initiative could redefine the very architecture of software companies. By erecting an autonomous system capable of coding, testing, designing, and interacting, Musk isn’t just playing the innovator — he’s building a self-sustaining AI civilization.
Who would have thought of a virtual Microsoft able to self-generate, self-optimize, and deploy without humans — a revolution with the flavor of science fiction, but all too real.

4️⃣ AI-boosted Pixel 10: Google sets the smartphone market on fire

The summary: Google has unveiled the Pixel 10 series along with a flood of AI-powered features designed to simplify daily life. From live call translation in your own AI-generated voice, to 100x Pro Res zoom, a photo coach, redesigned voicemail, and an intelligent journal — the Pixel ecosystem now packs spectacular upgrades. The Pixel Watch 4 and new Gemini integrations complete the lineup.

Details :

• Photo Coach: Scans your scene, suggests inspiring examples, and guides you step by step to capture the perfect shot.

• Magic Cue, the smart SMS assistant: Digs into your Google apps (Gmail, Calendar, Keep) to pull the right info and send it instantly — all stored locally.

• Pro Res zoom up to 100x: Exclusive to the Pixel 10 Pro, this mode fuses over 200 images with generative AI for a surprisingly sharp result.

• Natural language photo edits: Just say “remove the plastic bag” and the picture is fixed.

• AI voicemail: A “Take a message” option replaces the classic answering machine with transcriptions and smart reminders.

• Realistic voice translation: Calls in German, Japanese, or Spanish are translated live — in a credible imitation of your own voice.

• Guided digital journal: Writing suggestions, wellness tracking, and secure locking included.

• Daily Hub: A subtle dashboard showing appointments, reminders, and personalized recommendations.

• Pixel Watch 4 wrist-activated: One gesture launches Gemini for natural answers.

• Gemini Live with personality: Soon the assistant will speak fast, slow — or even like a pirate.

Why it's important: With these 10 new features, Google is turning the Pixel 10 into a portable AI lab. The smartphone evolves into a coach, secretary, translator, and even a director of your digital life. It’s a leap toward the “intelligent copilot” vision of the industry — though it may chip away at our autonomy

5️⃣ Halo X: the AI glasses that whisper the perfect reply

The summary: Harvard-born startup Halo is preparing to launch Halo X, a pair of smart glasses designed to record all your conversations, remember forgotten details, and feed you the right response at the right time.
Announced price: $249 for pre-orders. Founders AnhPhu Nguyen and Caine Ardayfio call it the first step toward “vibe thinking” — a permanent alliance between brain and AI

Details :

•  On-the-fly augmented memory: Halo X continuously transcribes conversations, recalls forgotten preferences, and helps you shine socially even when your memory falters.

• Vibe as second brain: For Ardayfio, “vibe thinking” means working hand in hand with AI agents capable of generating ten times more ideas and comebacks than a human alone.

• Always on, never discreet: Unlike Meta’s Ray-Bans, Halo glasses show no recording light. They film and listen all the time — a direct clash with consent laws.

• A prototype already sparking outrage: Their first hack? Modifying Meta glasses to instantly recognize faces, addresses, and employers — triggering an uproar over ethics and privacy.

• Science or cognitive dystopia? Researchers at Microsoft and Carnegie Mellon warn that relying on AI undermines memory and critical thinking — a phenomenon dubbed “cognitive offloading.”

Why it's important: Halo X cuts straight to the core of the tech dilemma: expand human intelligence, or outsource it to the point of atrophy. Behind the futuristic gadget lies an existential question: does AI make us smarter — or just more dependent?
If systems can design, test, and refine their own tools, innovation accelerates exponentially. Companies are redefining productivity itself and pushing the boundary between human input and machine autonomy.

❤️ Tool of the Week: Grok 2.5 Goes Open Source

Elon Musk just let go of the reins: Grok 2.5, xAI’s flagship model, is now available open source on Hugging Face! A thunderous announcement — with the promise that Grok 3 will follow in six months.

What’s it for?

• Peek under the hood: Download the model and see how it really works.

• Tinker at will: Customize Grok for your own apps, tests, or experimental projects.

• Contribute to research: Audit, improve, or invent new features through open collaboration.

• Test locally: Run it on your own machines to keep data in-house and experiment without the cloud.

• Mind the limits: The license is restrictive! No commercial use, no retraining of competing models.

• How to use it? Head to Hugging Face, where Grok 2.5 is freely downloadable. A decent GPU, a bit of elbow grease — and you can spin up your own Grok agents as if you had a mini-xAI on your laptop.