Welcome back! Uber just revealed that AI agents booked a grand total of … FOUR rides during its busiest week in October. For context, Uber and Uber Eats handled roughly 320 million rides and orders that same week, with AI agents responsible for just 500 food orders. Uber’s CEO says it’s still early days, which feels like a generous way to describe four customers. Apparently, the agentic revolution is moving fast everywhere except when it comes to getting a ride home.
AI Is Moving Off the Screen
This week I came across something that genuinely made me stop scrolling: Simulon 2.0.
I shared a clip of it, and the idea is simple but kind of wild. Instead of AI-generated stuff living inside a flat screen, it lives in the room with you:
You place virtual objects in the physical space around you.
You can actually interact with those objects, not just look at them.
It blends real-world surroundings with generated content in real time.
My POV: We spend a lot of time talking about AI in terms of chatbots, benchmarks, and who has the smartest model. But some of the most exciting stuff right now is happening where AI meets the physical world. Mixing what's real with what's generated is going to change how we play, learn, design, and shop, probably faster than most people expect.
Hit reply and let me know: if you could drop any virtual object into your living room right now, what would it be?
— Matt
Google It, or ChatGPT It?
Via CNBC
OpenAI revealed that ChatGPT now handles roughly one-fifth as many search-like queries as Google, a sign of just how quickly chatbots are changing the way people find information online.
By the numbers: ChatGPT consumer product lead Adam Fry shared the figure at an OpenAI event this week. The company now has 1.2 billion weekly active users, up 200 million in just two months. Google still processes vastly more searches, but ChatGPT’s reported volume is impressive for a product that launched only four years ago.
The fine print: OpenAI defines a search as a question that triggers a web search inside ChatGPT, but it hasn’t disclosed its exact query totals or the Google numbers used for comparison. The 20% figure is OpenAI’s estimate, not an independently verified market share statistic. Still, the trend is difficult to ignore, especially as Google increasingly incorporates AI Overviews and AI Mode into its own search experience.
The content problem: The shift is happening as websites—the ones from which chatbots pull answers—become increasingly protective of their content. Reddit is moving toward requiring users to log in, partly to limit unauthorized scraping, while The New York Times and other publishers recently sued Brave over allegations that it collected and distributed copyrighted material to AI companies. As AI-generated answers become more popular, publishers are questioning how much value they’re getting in return.
The bigger picture: We’re moving from searching for information to asking AI to find and synthesize it for us. Even Google is pushing users toward AI-generated answers instead of traditional links. But that creates a complicated question: what happens to the websites, publishers, and communities producing all that information if people stop visiting them? AI search needs a healthy internet to draw from, and the economics of that relationship are far from settled.
Apple Opens Siri AI to Third-Party Apps
Via Fortune
Apple announced that developers can now integrate Siri AI’s advanced capabilities into their own apps, allowing the assistant to perform more complex tasks without users constantly switching between applications.
What’s changing: The new developer preview, available through the iOS 27.3 beta and corresponding Apple software updates, expands Siri’s ability to understand personal context, recognize what’s happening on screen, and take actions inside third-party apps.
What Siri can do: Developers can use Apple’s App Intents framework to connect their apps to Siri and create more complex workflows. For example:
Ride-sharing: Ask Siri to change your pickup location while tracking an active ride, without opening the app.
Shopping and orders: Check deliveries, place orders, or book reservations through supported apps.
Multi-step tasks: Use new app skills to combine several actions into a single request.
A more personal assistant: Apple is also expanding Siri’s capabilities across its own apps, including Mail, Messages, Calendar, Photos, and Reminders. Users can ask Siri to find a coffee shop they visited last week or archive files they haven’t opened in over a year. The third-party integrations are currently in developer preview, with broader availability expected in the coming months.
Why it matters: AI assistants are moving beyond answering questions and toward actually doing things across the apps people already use. Apple has a major advantage here because Siri is built directly into its devices and operating systems. The more context assistants can understand and the more apps they can control, the less users need to think about which app to open.
Your App Doesn’t Need an Essay. It Needs a Decision.
Most tasks don’t require a long-winded AI response. Your app just needs to know which action to take. That’s where OpenAI’s Decisions API comes in. Give Decisions API context (text, images, or JSON), define your question, and set the allowed answers or scoring rubric. You get back focused answers.
Decisions API gives your app a choice, probability, or score it can actually use to:
Classify requests as billing, technical support, or account access.
Estimate whether a product has visible damage based on photos.
Evaluate severity against ordered levels, like cosmetic, workaround available, or fully blocked.
Choose among continuing, retrying, or asking for clarification in your agent workflows.
Classify business documents and route them to the next step.
Ready to stop micro-managing AI? Try one recurring app decision today.
Turn Your Product Into an Interactive Demo
Via AtriumKit
AtriumKit is a Chrome-based tool that records your interactions with a web app and turns them into interactive product demonstrations. It creates guided walkthroughs with tooltips and branching paths, as well as clickable sandboxes that let viewers explore products independently.
How you can use it
Record product demos by clicking through your web app
Create guided walkthroughs and interactive sandboxes
Generate AI voiceovers and multilingual captions
Embed demos in onboarding pages, help centers, or emails
Edit and analyze demos through AI tools like Claude and Cursor
Pricing: Free and paid
Give Your Browser a Persistent Memory
Via writenode
writenode is a Chrome extension that connects notes, highlights, saved pages, and research context across browsing sessions. It remembers information you’ve encountered before and surfaces relevant connections when you’re researching related topics, helping you avoid repeating the same searches or losing track of ideas.
How you can use it
Save notes, highlights, and webpages across sessions
Connect new research with previously saved information
Rediscover useful sources without searching again
Maintain context across multiple projects and browser tabs
Pick up research where you left off
Pricing: Free and paid
Jobs, announcements, and big ideas
Google Vids adds Omni 1.2 Flash for free, longer 15-second AI video clips.
Meta gives its Muse assistant a voice and expressive avatars.
Claude launches Personal Finance to answer money questions from your connected accounts.
OpenAI unveils GPT-6 Cyber, a model that hunts for exploits within authorized scope.
Runway explains how it runs real-time Characters at 24 fps on half the GPUs.
OpenAI rolls out Images 2.6 with GIF support and 50% faster generation.
Claude helps map the first 3D structure of a prokaryotic vault particle.
AI is finally getting good at editing videos. I go through the step-by-step of how Claude edits my videos.
That’s a wrap! See you this Friday for more.
