Good morning. Google spent the better part of I/O 2026 introducing a gazillion new AI products. Then the keynote drifted into full Burning Man philosophical mode. DeepMind CEO Demis Hassabis says we may be in the “foothills of the singularity” as AI ushers in a new golden age of scientific discovery and improves the lives of everyone, everywhere. Bold prophecy from the folks who made search worse and called it progress.
Google is building OpenClaw for Workspace. Spark is an always-on Gemini agent that runs across Gmail, Docs, Sheets, Slides, Chrome, and eventually your local files. Since it lives on Google Cloud, it can keep working after your laptop closes or your phone goes dark. Google says Spark will draft emails, dig facts from your docs, monitor your inbox, track your schedule, track your schedule, and keep the work flowing while you do literally anything else. It’s rolling out to trusted testers now, with beta access opening next week for Google AI Ultra subscribers. Now Google gets to find out if people hate office agents or just Microsoft’s version. (Google)
OpenAI is turning its compute shortage into a product. The new offering, called Guaranteed Capacity, lets customers lock down long term access to the compute power they need for AI. They can commit for one, two, or three years, with bigger discounts for longer terms. Sam says customers want to reserve their capacity now, because the world will be short on compute for a while. OpenAI says it will only sell the product until its current allocation runs out. If your AI roadmap depends on OpenAI’s servers, you can now pay for peace of mind. Funny how quickly a bottleneck becomes a business model. (OpenAI)
Capture live traffic, fine-tune and optimize, then deploy your own checkpoints to dedicated GPU endpoints.
Choose hardware, set scaling limits, and select region. Stable latency, predictable cost, clear data residency.
👀 closer look
OpenAI slashed its $1.4 trillion Stargate investment to $600 billion. Remember when Sam stood at the White House and talked about Stargate like the bulldozers were already warming up? 16 months later, OpenAI still hasn’t hired staff or broken ground on its own data centers. Its UK project was paused, the Norway site shifted to Microsoft, and the Texas expansion with Oracle got canceled. OpenAI still needs infinite compute, so what’s really going on? The cleaner IPO story is staring everyone in the face. Building data centers makes OpenAI look like an expensive infrastructure company. Renting compute makes it look like software while somebody else carries the debt. Convenient how the bulldozers vanished right before Wall Street got its tour. (Tech Times)
Companies are drowning in data and still miss what matters. Brands need to know what customers care about right now. The answer is usually scattered across social media, news, reviews, patents, search behavior, and market data. Quid pulls all these signals into one place using advanced AI trained on more than 2 petabytes of active data. Teams at Walmart, Hyundai, Intel, Accenture, NASA and The Microdose AI use it to spot high signal trends every day and quickly act on what deserves attention. Quid turns signals into business decisions, showing what’s changing, why it matters, and what’s coming next. Want to take it for a spin? Quid is now offering a free trial.
Scientists want to resurrect the dodo, but first they have to reinvent the egg. Colossal Biosciences just hatched 26 chicks from 3D printed eggshells to learn what it takes to bring back extinct birds. The problem comes after the DNA work because the embryo still needs somewhere to develop. The giant moa makes that obvious. Its egg would be 80x the size of a chicken egg. Colossal’s idea is to move the embryo into a larger artificial egg before the real shell runs out of room. Other scientists have grown birds in artificial containers before, so Colossal didn’t invent the category. They just printed a better incubator. (MIT)
fun stats
♟️ 2.5 years. Total time “vibe coding” inventor Andrej Karpathy spent at OpenAI across two stints before bouncing to Tesla, leaving again, launching Eureka Labs, and now joining Anthropic.
🟡 $300 million. What Salesforce expects to pay Anthropic for coding tokens this year. That’s roughly 4.5% of Salesforce’s total outside tech supplier bill from last year.
🛍️ 32%. Increased chance consumers buy a product after reading an AI generated positive review summary, even though it hallucinated follow up answers.
🔥 3.2 quadrillion. Monthly AI tokens now consumed by Google users, up from 480 trillion a year ago. Everyone, everywhere is tokenmaxxing.