In partnership with
Welcome to Jumble, your go-to source for AI news updates. This week, Google I/O is two weeks away and it’s teasing what may be the biggest Gemini reveal of the year. Meanwhile, Yann LeCun just raised over a billion dollars to build something that explicitly is not a chatbot. Let’s dive in ⬇️
In today’s newsletter:
📅 Sundar Pichai previews I/O 2026
💸 Yann LeCun’s big bet against LLMs
👔 Sierra hits $15.8B in record-breaking raise
⚔️ Pentagon signs AI giants, shuts out Anthropic
🧪 Weekly Challenge: Stop your AI from making things up
🎤 Google I/O Is Two Weeks Out and Pichai Is Hyping Gemini 4
Google I/O 2026, Google’s annual developer conference, kicks off May 19 in Mountain View, and on the Q1 earnings call Sundar Pichai promised Google is pushing the next frontiers of agents and agentic coding.
🕶️ Smart Glasses and Aluminum OS Lurk in the Wings
Samsung's Android XR smart glasses are reportedly close to launch at $379 to $499. A formal reveal of Aluminum OS, the long-rumored ChromeOS-Android merger, would also fit the moment.
🤖 Gemini 4 Is the Reveal Everyone Wants
With Gemma 4’s recent release (Google’s open source AI), industry watchers are expecting a Gemini 4 reveal alongside updates to Veo, Project Astra, and Gemma.
Pichai also confirmed AI Overview costs have dropped 30% since the move to Gemini 3.
What do you most want to see at I/O 2026?
🛍️ Plus a Huge TPU Move
Pichai also announced Google will start selling TPUs directly to outside customers for in-house data centers. That's a direct shot at NVIDIA's near-monopoly on AI training hardware.
“AI is Going to Fundamentally Change…Everything”
That’s what NVIDIA CEO Jensen Huang just said about the AI boom, even calling it “the largest infrastructure buildout in human history.”
NVIDIA’s chips made this real-time revolution possible, but now it’s collaborating with Miso to unlock amazing new advances in robotics.
Already a first-mover in the $1T fast-food industry, Miso’s AI-powered Flippy Fry Station robots have worked 200K+ hours for leading brands like White Castle, just surpassing 5M+ baskets of fried food.
And this latest NVIDIA collaboration unlocks up to 35% faster performance for Miso’s robots, which can cook perfect fried foods 24/7. In an industry experiencing 144% labor turnover, where speed is key, those gains can be game-changing.
There are 100K+ US fast-food locations in desperate need, a $4B/year revenue opportunity for Miso. And you can invest alongside industry powerhouse Ecolab and become a Miso shareholder today to unlock up to 7% bonus stock. But hurry, this bonus stock offer changes in only a few days on May 7. Invest in Miso now.
This is a paid advertisement for Miso Robotics’ Regulation A offering. Please read the offering circular at invest.misorobotics.com.
🧠 Yann LeCun Just Bet $1 Billion Against Large Language Models
Meta's former chief AI scientist just raised $1.03 billion for AMI Labs, his Paris-based startup building AI without large language models.
The four-month-old company is now valued at $3.5 billion, making it the largest seed round in European history.
🚫 LeCun Thinks LLMs Are a Dead End
LeCun has spent years arguing predicting tokens isn't real intelligence. AMI is pursuing world models that learn how reality actually behaves, not just what word probably comes next in a sentence.
🌍 The Money Is Following Him
Bezos Expeditions, NVIDIA, Samsung, Eric Schmidt and Mark Cuban all wrote checks. AMI's first partner is healthcare AI startup Nabla, where hallucinations can have life-threatening consequences for patients.
Weekly Scoop 🍦
🎯 Weekly Challenge: Train Your AI to Stop Lying
Challenge: AI hallucinations are the number one reason your boss still doesn't trust the report you generated last Tuesday. The fix isn't a better model, it's better prompting. This week, lock down four habits that cut hallucination rates by up to 36%.
Here’s what to do:
📚 Step 1: Make it cite. Add "cite authoritative sources for each claim" to factual prompts. Anchored references mean fewer invented stats and easier verification on your end.
🔁 Step 2: Force a chain of thought. Tell your AI to "reason step-by-step before answering" to cut logic gaps and stop mid-sentence fabrication on multi-part questions.
📎 Step 3: Feed it the source. Paste the document or data directly into the prompt. Anthropic's own guidance says grounding answers in real text dramatically cuts invention.
🤷 Step 4: Let it bail. Add "if you are not certain, say so" to your system prompt. Models confabulate because they think they have to answer. Permission to say "I don't know" is the simplest fix in the book.
What does Google have to drop at I/O to actually move the needle? And will LeCun's bet pay off, or is he the world's most expensive contrarian? See you next time! 🚀
Stay informed, stay curious, and stay ahead with Jumble!
Zoe from Jumble
