Sam Altman Wants to Scan Your Eyeballs

Welcome back apprentices! 👋

You ever looked at a glowing orb and thought, “Yeah, this thing might change the internet?” 

No? Well, Sam Altman has — and now people are lining up to stare into one.

In a world where AI can fake your voice, face, and “sincere apology for the inconvenience,” proving you’re a real human is about to become a hot business necessity.
Forget CAPTCHA boxes and email codes — this next-level tech wants your eyeballs. Literally.

What’s behind it? Who’s already signed up? And why are Visa and Tinder involved? Let’s just say… the future of identity is blinking back at you.

In today's email

• Duolingo Just Replaced Its Human Tutors 

• AI Leaderboards Are Kind of Rigged

• Google’s Gemini Is Now Chatting With Kids

• Scanning Eyeballs to Fight AI Bots 

• Meta’s New AI App

• Even more AI magic

• Test the prompt

_{Read Time: 4 minutes}

Quick News 

👋 Adiós, Contractors! Duolingo’s gone “AI-first,” and it’s not just a vibe shift — it quietly swapped out a bunch of its contract translators and content creators for generative AI. Why? Because GPT-4 can now churn out grammar drills and vocabulary quizzes faster than you can say “Dónde está la biblioteca?” The company says it's freeing up humans for more “creative” work... which we assume doesn’t include making flashcards. This is a loud wake-up call for anyone in content-heavy industries.

🏆 A new study just dropped the equivalent of a mic on Chatbot Arena — the leaderboard where top AI models duke it out for bragging rights. Turns out? The fight isn’t fair. Major players like Meta, Google, and OpenAI are quietly testing tons of private model variants, cherry-picking the best, and soaking up nearly 40% of all user feedback data. Meanwhile, open-source models are ghosted, under-sampled, or silently booted from the leaderboard. It’s like entering a bake-off where some contestants submit 27 cakes, then only serve the winner.

🧒 Starting this week, Google will let kids under 13 use its Gemini chatbot — yes, the one that sometimes still fumbles basic facts. Access comes via Family Link, with extra safety guardrails and a strict “we won’t use your kid’s chats to train AI” policy (thanks, Google). It’s part of the big-tech race to win over the next generation before they can spell “algorithm” — while UNESCO is somewhere in the background, waving regulatory flags and whispering, "maybe not yet?"

Sam Altman 
Altman’s Orb Is Scanning Eyeballs to Fight AI Bots (No, This Isn’t Sci-Fi)

Sam Altman — yes, the ChatGPT guy — just launched six U.S. retail locations where you can get your eyeballs scanned by a shiny chrome sphere called the Orb. Why? Because his startup “World” (formerly Worldcoin) wants to prove you’re human in a world full of increasingly convincing AI-generated content and bots.

You step up, stare into the orb for 30 seconds, and boom — you're issued a unique “IrisCode” and some free crypto. Verified humans can then use their World ID to log into places like Shopify, Discord, Reddit, or even Tinder Japan.

 It’s all part of a plan to build a global proof-of-personhood network — because apparently CAPTCHA isn’t cutting it anymore.

What’s It Good For (Besides Scanning Your Soul)?

• 🔐 Bot-proof login: Use World ID to access platforms like Telegram, Minecraft, and Reddit — no more “forgot password” drama.
  
  

• 💳 Payments & perks: Visa’s rolling out a “World Visa card” — but only if your iris gets a thumbs-up.
  
  

• ❤️ AI-proof dating: Match Group is testing World ID on Tinder to keep bots (and catfishers) off your love life.
  
  

• 🌍 Global identity layer: Your AI-era passport — minus the customs line and paperwork.
  
  

• 🪙 Bonus crypto: Get free WLD tokens just for signing up — because nothing says “we appreciate your eyeballs” like blockchain coins.

Why It Belongs in Your AI Feed

This is not a crypto gimmick — it’s a real attempt to solve the looming question:

 “How do we know who’s human when AI can fake everything?”

From fake tweets to deepfake videos, LLM-written scams to identity fraud, platforms are scrambling to keep trust intact. 

World’s bet? That biometric, decentralized identity becomes the firewall between us and synthetic chaos.

And if your business relies on real people — customers, clients, or community — start paying attention to digital ID tools like this. AI is making it harder to tell who’s real online, and trust is becoming a competitive advantage. You don’t need to scan eyeballs yet, but you do need to think about how your business will stay human in an AI-first world.

(Also, if you're still using “What’s your mother’s maiden name?” as verification… time to upgrade.)

 We’ll share the results in Friday’s issue — and honestly, we’re dying to know.

Meta
Meta’s New AI App Might Be Your Next Hire

Meta just launched its very own AI assistant app, powered by Llama 4 — and it's not just for tech bros and Zuck’s smart glasses. It talks, listens, remembers what you like, and runs across Meta’s ecosystem (Instagram, Facebook, Messenger, Ray-Bans, and the web).

With voice support, image generation, memory, and a trending prompt feed, this is Meta’s full-stack assistant coming for your inbox, social feed, and maybe even your sunglasses.

What’s It Actually Good For?

• 🎙️ Talk While You Hustle
  Create lists, answer emails, or plan a launch — just by speaking.
  
  

• 💡 Instant Post Ideas
  Ask for a caption. Get something witty. Post like a pro.
  
  

• 📄 Skip the Fine Print
  Upload a doc. Get the gist. Ignore the jargon.
  
  

• 📱 Smarter Content Help
  Link Instagram or Facebook for tailored ideas and replies.
  
  

• 🕶️ Voice AI in Your Shades
  Talk to your glasses. Ask anything. Look cool doing it.
  
  

• 📊 Trend Radar On
  Use “Discover” to see what others are asking. Stay ahead.
  
  

• 🧠 Always-On Brainstorm Buddy
  Need names, replies, or stats? It’s ready. And drama-free.

So What’s the Deal?

This isn’t just a chatbot. It’s Meta trying to wrap its AI arms around your entire digital life — from your social DMs to your sunglasses. It's fast, voice-friendly, and social-media fluent, which means it's basically auditioning to be your AI co-worker, assistant, and sidekick — all in one.

And because it remembers things you’ve said and works across platforms, it’s becoming something your customers might expect too. Imagine AI that can schedule meetings, summarize messages, and help you write marketing copy... without sounding like a robot from 2011.

Try it today. Ask for a post idea, a content plan, or help writing a reply.

It’s the AI intern you didn’t hire — but somehow, it already knows your workflow.

Help Your Friends Level Up! 🔥

Hey, you didn’t get all this info for nothing — share it! If you know someone who’s diving into AI, help them stay in the loop with this week’s updates.

Sharing is a win-win! Send this to a friend who’s all about tech, and you’ll win a little surprise 👀

Even Quicker News 

🏠 Airbnb quietly launched an AI support rep now handling 50% of U.S. customer issues — like a tireless, hyper-polite superhost with no feelings and no PTO. It’s fast, unfazed, and never puts you on hold… unless it thinks you need a moment.

🏷️ Pinterest is now tagging AI-generated images with “AI modified” labels after users noticed suspiciously flawless content. Behind the scenes, smart tools scan metadata and visuals to spot even subtle fakes.

🚗 Waymo and Toyota are teaming up to put self-driving smarts into personal cars — because why should robo-taxis have all the fun? Soon, your next Toyota might come with an AI that drives better than your cousin Dave. 

Today’s Toolbox

🧮 A new AI model just dropped, and it’s basically your smartest coworker — if your coworker was a calculator that read every math book ever and never needed coffee breaks. It can solve super complex problems, prove the answers are correct, and explain its logic — all without making a single Excel error.

DeepSeek-Prover V2 is already outperforming top models (yes, even GPT-4) on elite math benchmarks like MiniF2F and Putnam. But here’s the kicker: it’s open-source, which means it’s not just for scientists in lab coats — it’s for you too.

What Can It Actually Do for You?

You’re not solving theorems all day — but you are juggling numbers, logic, and decisions. Here's how DeepSeek-Prover–style AI can actually help in your workday:

• 🧾 Spot sneaky spreadsheet errors before your budget presentation goes sideways.
  
  

• 📋 Help build rule-based systems, like pricing models or scheduling logic.
  
  

• 🛠️ Verify custom tools in finance, construction, supply chains, or operations that rely on formulas or logic trees.
  
  

• 🎓 Create explain-it-like-I’m-five training tools for employees learning technical tasks.
  
  

• 💬 Build smarter AI assistants that don’t just “say stuff” — they can actually prove it’s right.
  

Basically, if it’s something you’ve ever double-checked twice, this model could check it for you — and not screw it up.

Why It Matters to Your Business

This AI isn’t here to replace jobs — it’s here to make you look brilliant without breaking a sweat. It brings serious logic skills to the table, but unlike your high school math teacher, it’s helpful, polite, and available 24/7.

In a world where AI sometimes makes things up (hi hallucinations), DeepSeek-Prover says: “Hold my theorem.”
It’s part of a movement toward AI that isn’t just confident — it’s correct.

Start thinking: Where in my business do I need logic, rules, or things to be provably correct?

Whether it’s pricing rules, workflow automations, or building smarter AI tools, this model is a sign of what’s coming: AI that doesn’t just “sound smart” — it is smart. And provably so.

Try it, test it, or build on top of it — before someone else builds it into the product your customers leave you for.

🧪 Test the Prompt

A playground for your imagination (and low-key prompt skills).

Each send, we give you a customizable DALL·E prompt inspired by a real-world use case — something that could help you in your business or job if you wanted to use it that way. But it’s also just a fun creative experiment.

You tweak it, run it, and send us your favorite. We pick one winner to feature in the next issue.

Bonus: you’re secretly getting better at prompt design. 🤫

👑 The winner is…

Last week, we challenged you to test GPT-4o’s visual generation skills with this prompt.

Congrats to XX from XX!🥳

Want to be featured next? Keep those generations coming!

🎨 Prompt: The Mindset Vending Machine

“Design a futuristic vending machine that dispenses mindsets instead of snacks. The exterior is styled in the aesthetic of [cultural or time period influence], and its display screen uses icons shaped like [abstract or symbolic object] to represent different mental states. A user stands in front, deciding whether to select “Focus,” “Confidence,” or “Let It Burn.”
Style: ultra-sharp product concept art with neon glow, reflections, and soft shadows.”

We’ll be featuring the best generations in our next newsletter!

_FEEDBACK