This week, we’ve got 5 incredible AI breakthroughs and developing stories; from a quiet stock market correction to a the leaderboard fight between xAI and Google. Then, we zoom into the new "Codex Pro" agent that might just code you out of a job (or at least a weekend shift). Let’s dive in ⬇️

In today’s newsletter ↓
🕵️ 5 AI news stories that deserve a second look
🦾 ChatGPT 5.1 Pro & Codex Pro go agentic
🤝 Softbank teams up with OpenAI
📜 EU lightens AI rules for companies
🗣️ Weekly Challenge: Learn how to use ChatGPT group chat

Here are five stories that mattered more than their headlines suggested. From a new Gemini 3 model topping charts to a stock market wobble that hints at bubble trouble. Then, we learn all about GPT 5.1 Pro and Codex Max. Let’s dive in! ⬇️

While everyone was laughing at its "Elon vs. LeBron" takes, xAI quietly released Grok 4.1, and the benchmarks are no joke. The new model features a "Thinking" mode that has shot to the top of the LMArena leaderboards, boasting a massive jump in emotional intelligence (EQ-Bench) and creative writing.

xAI claims it is now three times less likely to hallucinate than previous versions, positioning it as a serious contender for "smartest model" rather than just the "anti-woke" alternative.

Google didn’t just launch a model; they overhauled their entire search infrastructure. Gemini 3 is now live, and unlike previous iterations, it’s built for agentic workflows—meaning it can handle multi-step tasks across your Google apps (Calendar, Gmail, Maps) without losing the thread.

It’s the "brain" behind the new Nano Banana Pro, but its real power is in its deep reasoning capabilities, which are now being deployed to answer complex queries directly in the search bar.

Yes, the name is silly, but the tech is enterprise-grade. Google’s Nano Banana Pro (officially Gemini 3 Pro Image) solves the "consistency crisis" in AI art by allowing Multi-Reference Fusion with up to 14 input images.

This means brands can finally generate consistent characters and products across different scenes. It also introduces "Search Grounding," forcing the model to check real-world geography and data before rendering, making it a killer tool for accurate infographics.

OpenAI just made AI social. The new Group Chat feature has rolled out globally, allowing you to invite friends and family into a shared space with ChatGPT. Crucially, OpenAI has walled this off from your private data—individual memories are not shared in the group context.

It’s designed for collaborative tasks like trip planning or settling debates, effectively turning ChatGPT into a neutral, super-smart moderator for your group chat drama.

While tech enthusiasts cheered for new models, Wall Street got cold feet. Nvidia and other AI stocks took a sharp dip this week despite strong earnings, driven by renewed fears of an "AI Bubble."

Investors are increasingly skeptical about the massive infrastructure spending by Meta, Microsoft, and Google, asking when the trillion-dollar investment will turn into trillion-dollar profits. This "correction" signals a shift from blind hype to a demand for tangible ROI.

They’re both investors in AMASS Brands. And you can join them. Why invest? Their rapid growth spans everything from organic wine to protein seltzers. So with consumers prioritizing health in the $900B beverage market, it’s no surprise AMASS earned $80M+ already, including 1,000% year-over-year growth. They even reserved the Nasdaq ticker $AMSS. Become an AMASS shareholder and secure limited-time bonus stock by Dec. 4.

Invest in AMASS Brands Now

_{This is a paid advertisement for AMASS’s Regulation CF offering. Please read the offering circular at https://invest.amassbrands.com}

While the world was distracted by image generators, OpenAI dropped a massive update for the builders. The new GPT-5.1 Pro and Codex Pro (technically GPT-5.1-Codex-Max) aren't just faster; they are fundamentally more autonomous.

The headline feature for developers is Compaction. This new technology allows the model to work on long-horizon tasks for over 24 hours. Previously, if you asked an AI to refactor a huge codebase, it would run out of "context" (memory) halfway through.

Compaction allows the model to intelligently "prune" its own memory, keeping only the relevant context while discarding the fluff. This means it can essentially run forever, iterating on code, fixing bugs, and testing its own work until the job is done.

Giving an AI autonomous control over your terminal sounds terrifying, and OpenAI knows it. That’s why Codex Pro operates in a strict default sandbox, with network access disabled unless explicitly granted.

It creates a "containment zone" for the agent to break things safely, ensuring that a runaway refactor loop doesn't accidentally delete your production database or leak API keys to the open web.

For the first time, OpenAI has also trained this model sto work with Windows environments, expanding its utility beyond the Mac/Linux-dominated dev world. Plus, it’s roughly 30% more token-efficient, meaning you get more "smart" for your dollar.

This is the shift from "AI as a Copilot" (which suggests code as you type) to "AI as an Agent" (which you hire to go fix a bug while you sleep).

🔋SoftBank scraps EVs to build OpenAI's physical backbone

🛑 Europe hits the snooze button on its strict AI rulebook

📱 Perplexity puts an agent in your pocket with "Comet" for Android

🤝 Cisco backs the "Godmother of AI" to give robots spatial awareness

🤖 Elon's latest forecast: Poverty ends, work becomes optional

📉 Bubble trouble? Markets slide despite Nvidia's strong numbers

Collaborating with AI is usually a solo sport. This week, let’s make it a team effort. We’re testing the new ChatGPT Group Chat feature to see if it can actually organize humans.

Here’s what to do:

📱 Create the Space: Open ChatGPT (mobile or web) and tap the

"People" icon (top right) to start a new Group Chat.

🔗 Invite the Squad: Send the invite link to 2-3 friends or colleagues. (Note: They need a ChatGPT account).

💬 The Prompt: Tag @ChatGPT and ask it to: "Plan a dinner itinerary for us this Friday. We need a cuisine that suits [Friend A]'s vegetarian diet and [Friend B]'s love for spicy food, within a 15-minute drive of downtown."

🗣️ Iterate: Have your friends reply directly to the AI with their objections or preferences (e.g., "No Italian food"). Watch how it negotiates the conflicting constraints.

Score It: Did it find a solution faster than your usual 40-message text thread? If yes, you just upgraded your social life.

That’s it for this week! Were you surprised by the number of game-changing AI updates this week? And, can OpenAI stop Google’s momentum? Hit reply and let us know your thoughts.

Zoe from Overclocked