Welcome back apprentices! 👋

Ever feel like your weather app is just flipping a coin? 

One day it says sun, next day it rains — it's basically astrology with a radar. But what if forecasts stopped guessing and started knowing? 

A new AI model just entered the chat, and it might be the reason your umbrella finally stays folded when it's supposed to. Let’s just say... the days of 50/50 forecasting could be numbered.

And yes — this one predicts tropical cyclones from a laptop, anywhere in the world.

In today's email

• Siri’s New Boss

• TikTok, But Make It Smart

• Meta AI Lands in Europe

• AI Stock Watch

• AI Forecasts Are In…

• NVIDIA’s Grid Gameplan

• And yes, even more AI magic

_{Read Time: 4 minutes}

Quick News 

🍏 Apple just hit Ctrl+Alt+Delete on its AI leadership. Vision Pro legend Mike Rockwell is now steering the Siri ship after internal chaos, missed deadlines, and AI features so delayed, even Siri couldn’t find them. He’ll report directly to Craig Federighi, effectively ghosting current AI lead John Giannandrea. With personalization upgrades and Siri 2.0 still stuck in beta limbo, Apple is hoping a leadership makeover can stop it from falling even further behind OpenAI, Google, and... literally everyone else.

📱 In what might be the world’s most ambitious makeover, AI search startup Perplexity just pitched a plan to rebuild TikTok from the inside out — and yes, it includes citations. The company wants to replace TikTok’s mysterious “For You” algorithm with an open-source, privacy-friendly, American-made version powered by Nvidia’s new Dynamo chips. The goal? Make TikTok more like Wikipedia and less like a slot machine — with videos that teach you stuff, not just hijack your dopamine.

They’re proposing real-time fact-checking, AI-enhanced search, and even multilingual video annotations — so yes, your next cat video could come with footnotes and a translation in 17 languages. If Perplexity pulls it off, TikTok won’t just entertain — it’ll explain.

🎉 After fashionably missing the invite for over a year, Meta AI has officially arrived in 41 European countries (plus 21 vacation spots) — sliding into your DMs on Facebook, Instagram, WhatsApp, and Messenger with a shiny blue icon and multilingual charm. This chatty assistant already has 700M+ monthly users worldwide and now wants to help you plan holidays, settle group chat feuds, or figure out what the heck “smart casual” means in six European languages. 

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 let’s bring them into the fold!

Aardvark
The AI That Can Predict a Cyclone on a Laptop

Cambridge University, the Alan Turing Institute, and Microsoft just teamed up to drop Aardvark — not an animal, but a lightning-fast, fully AI-driven weather prediction model that could seriously shake up the forecast industry. 

Unlike traditional models that gobble terabytes of data and require massive supercomputers, Aardvark runs on a desktop with just 10% of the data, but still outperforms the U.S. GFS model on key metrics like temperature, wind, and precipitation.

🔍 Aardvark vs. GFS vs. GenCast

• Forecast Speed: Aardvark produces a 10-day forecast in just 5 minutes — 72x faster than GenCast and 72x faster than traditional GFS (~6 hours).

• Hardware & Accessibility: While GFS and GenCast require supercomputers or GPU clusters, Aardvark runs on a desktop.

• Carbon Footprint: Lower compute needs = significantly lower emissions — Aardvark is greener by design.

Trained on 39 years of climate data from NASA and ECMWF, Aardvark uses deep learning and autoregressive transformers to generate five-day global forecasts in mere minutes. The model shows particular strength in hard-to-predict equatorial zones and even handles extreme weather events. It’s also open-source, low-emission, and highly scalable — which could democratize weather forecasting for regions without access to traditional infrastructure.

What’s the deal? Aardvark may be the first weather model that’s as useful to climate scientists as it is to countries with no supercomputers. It’s a potential game-changer for regions in the Global South, offering a plug-and-play solution to build bespoke, locally-run forecasts. And if it scales up as planned, it could also displace traditional systems in the developed world by drastically cutting costs, energy use, and forecasting time. 

Bonus: it may finally democratize access to early warnings for floods, droughts, and hurricanes — just when climate chaos makes that more essential than ever.

☔ The next time you complain about a missed rain alert — remember, Aardvark saw that storm coming while sipping coffee on a MacBook.

Halfway through today’s edition — how are we doing?

Take 10 seconds and let us know what you love, what could be better, or any brilliant ideas you’ve got!

Now, let’s dive back in — what’s AI been up to this week?

NVIDIA
When AI Needs More Power… and Becomes the Power Solution

NVIDIA and EPRI just plugged into one of the most ironic AI challenges yet: AI is draining the grid — and now it’s tasked with saving it. Their new Open Power AI Consortium brings together a powerhouse crew of utilities (like Duke, PG&E, and Con Edison) and tech giants (Microsoft, Oracle) to build open, domain-specific AI models that can supercharge how we plan, manage, and distribute electricity.

With global electricity demand expected to rise 4% annually through 2027 — nearly double the growth seen in 2023 — and data centers multiplying like caffeinated gremlins, the energy sector can’t afford to wait. 

NVIDIA’s H100s are being used to train grid-optimizing AI agents, and the results will roll out as NIM microservices that plug straight into operations.

Old Grid vs. New Grid, NVIDIA Edition:

• 🔌 Traditional Grid: Takes 4 years to run feasibility studies
   🤖 NVIDIA Grid: AI shaves that down by 5x — like the express checkout lane for electricity
  

• 🧾 Old Way: Regulatory paperwork generated by committees and coffee
   🧠 New Way: AI reads thousands of permits and writes your reports in hours (and doesn't need a lunch break)
  

• 🛠️ Traditional Scaling: Slow, reactive, and mostly Excel
  ⚡ New Scaling: Predictive maintenance and AI load balancing that plans ahead like your Type-A friend

• 💸 Old Infrastructure: Supercomputers, clunky software, and expensive grid bottlenecks
  💻 New System: Open-source models running on NIM microservices — ready to plug & play like Lego for energy geeks

• ⚖️ Peak Demand = Stress (and sometimes blackouts)
   🌞 AI Optimization: Unlocks 76 GW of spare capacity just by shifting workloads when the grid’s less cranky
  

So what’s the deal? AI is creating data center cities the size of small nations — and suddenly electricity isn’t just a utility bill, it’s a strategic weapon. 

Microsoft is hoarding solar like it's the new gold (backing 10.5 GW in clean energy), while NVIDIA is making sure we don’t accidentally blackout Earth training GPT-6.

With the Open Power AI Consortium, the grid finally gets an upgrade that isn’t just "more wires." We're talking foundation models trained on actual grid problems, deployable to utilities from Kentucky to Kenya. 

And the best part? They're open source. That’s right — you can finally fix your city’s blackout schedule without needing Elon’s phone number. And if your lights flicker mid-Zoom, don’t panic — NVIDIA’s grid bot is probably debugging your transformer.

Even Quicker News 

🧾 Microsoft wants to teach its AI models how to say “thank you” — preferably with royalties. It’s like ancestry.com for training data, except your memes and blog posts might finally pay rent.

🗺️ AI agents are tired of squinting at screenshots like it’s Where’s Waldo for robots — so Browser Use raised $17M to teach them how to actually read websites. Basically, it’s like giving your digital assistant a pair of glasses and a map.

💬 Meta's testing AI-generated Instagram comments, so your friend’s “🔥 fit” might soon come from a chip, not a chum. Authenticity is dead — long live the algorithm!

Today’s Toolbox

📡 Claude just got Wi-Fi! Anthropic finally gave its star pupil internet access — now it can Google stuff like the rest of us and actually cite its sources (take notes, ChatGPT)

🗣️ OpenAI just gave its bots a voice upgrade — now they can talk like pirates, poets, or bedtime storytellers on command. It’s fun and impressive, but let’s be real: ElevenLabs still sounds more human (and less like Siri’s weird cousin).

📬 Gmail’s new AI-powered search finally helps you find that one email from your boss buried under 47 promos and your aunt’s cat memes. It’s like Ctrl+F — but with a brain. 

Let's Chat About AI

This is all for today!

Replay to this email with your thoughts.

AI is more than just a buzzword. It’s a shift in how we live and work. And understanding it a bit better means you can make smarter choices about the tech you use every day.

_FEEDBACK