Welcome back! Elon Musk says The Boring Company is working on “a simple precursor Hyperloop” between Austin and San Antonio that could cut the trip to under 30 minutes.
This would be very exciting if The Boring Company did not have a long history of announcing tunnel projects that quietly disappear into the same place as my New Year’s resolutions. Only time will tell where these initiatives go!
Could AI Finally Save Us From Hold Music?
Two of the fastest-moving AI agents both shipped phone-calling features within hours of each other last week. Instinct launched a service called Concierge, and Meta's Muse rolled out outbound calls to US businesses. Both work roughly the same way—you tell your AI agent what you need, and it dials the business, handles the conversation, and reports back with a transcript and summary.
Google is moving in the same direction with Gemini-powered calling in Search. You ask Google to call a local business, and it'll get you the pricing and availability without you ever picking up the phone yourself.
These agents solves one of the biggest wastes of time most people deal with regularly: waiting on hold.
Here's the wrinkle worth knowing about, though. This week, 404 Media reported that Meta had been quietly routing many of the "AI" Muse calls to human contractors in a call center.
A Meta VP acknowledged this was "a miss" and said the company has rolled that back for now, but the reason it was happening in the first place is worth understanding: internal tests reportedly showed human callers hitting a 95-98% success rate on these calls, while the pure-AI version landed noticeably lower.
Which is a pretty good reminder that the tech isn't quite as ready as the marketing suggests. Yet…
My POV: There's a slight worry in the back of my mind about what this kind of technology means for spam calls and scams. Making automated conversations more convincing is going to get messy fast on the receiving end. Imagine being a small business owner fielding AI-placed calls all day, unable to tell which are real customers and which are attempts to social-engineer information out of you.
But honestly? I'm pretty excited to never have to waste three hours on a call with my insurance company where half is spent on hold, and the other half is spent explaining the same problem to three different reps.
Hit reply and let me know: what's the first call you'd hand off to your AI agent?
— Matt
Google Opens Googlebook Preorders
Via WIRED
Google opened preorders for Googlebook, a new premium laptop category built around Gemini and tighter Android integration.
What’s launching: Googlebook starts at $899, with initial models from Acer, Asus, Dell, HP, and Lenovo. The laptops will run on Intel or Qualcomm processors, offer up to 14 hours of battery life, and pair ChromeOS desktop foundations with the Android technology stack.
Gemini at the center: Google is positioning Googlebook as a laptop built for an AI-first workflow, not just a Chromebook with a nicer spec sheet. Gemini tools will help users draft and organize text, ask questions about what’s on their screen, and resume tasks from an Android phone.
Google wants the laptop to feel like part of the same AI-powered ecosystem as your phone, not a separate device you occasionally sync. The idea is to make moving between an Android phone and a laptop feel much more seamless.
Competitive angle: The pricing is interesting. Google is pushing upmarket with an $899 laptop line just as Apple has moved downmarket with the $699 MacBook Neo. That puts both companies closer to the middle: Apple trying to win more budget-conscious buyers, and Google trying to convince premium Android users they deserve a laptop as polished as a MacBook-iPhone setup.
The bigger picture: The laptop is becoming another front in the AI ecosystem race. Google, Apple, Microsoft, and Meta are all trying to put hardware, software, assistants, and personal context under one roof. The question is whether an all-in-one AI experience will be compelling enough to make people switch devices.
Meta’s Muse Surges Past ChatGPT
Meta’s Muse AI personal agent climbed to the top of the US iOS free app store less than two weeks after launch, briefly overtaking ChatGPT and beating rival AI apps like Claude and Grok.
Fast start: Muse reached 730,000 downloads in roughly the first five days after its Sept.8 debut and more than 2.5 million downloads by Monday, Sensor Tower said.
What Muse does: Meta is pitching Muse as a personal AI agent that can manage tasks across the web, including filling out forms, organizing inboxes, and taking action on users’ behalf. The app is powered by Meta’s Muse Spark family of models and is available for free, with paid subscription tiers at $20 or $100 per month depending on usage.
Agent friction: Muse is also running into the messy reality of agents acting on the internet. Amazon blocked Muse from accessing its shopping site, citing privacy and security risks, while Shopify partnered with Meta to allow agentic checkout in online stores.
The bigger picture: Muse fits directly into Meta’s broader superintelligence push. The company is not just trying to build smarter models; it wants AI agents that normal people actually use. If Muse keeps gaining traction, Meta could turn its consumer scale into a serious advantage in the agent race.
Stop Burning Credits on AI Video That Almost Works
You know the feeling: the prompt is close, the camera move is wrong, and now you’re spending more credits trying to fix a video you can’t properly preview. Artlist’s Seedance 2.5 Draft Mode solves the preview problem.
With its low-cost previewing, Seedance 2.5 lets you:
Generate a fast 480p draft to see how your prompt, subjects, camera movements, pacing, and reference assets work together.
Adjust as necessary and regenerate until you have your desired result.
Compare all of your concepts before committing to a finalized direction.
Render your approved version in 1080p.
Ready to experiment more, waste credits less? Start testing concepts today.
Text an AI Agent Through iMessage
Via Iris
Iris is a personal AI agent that works through iMessage and connects to more than 1,000 apps. Users can text what they need, and the agent pulls context from tools like Gmail, Calendar, and Drive to complete tasks without repeated explanation.
How you can use it
Delegate tasks by texting an AI agent
Build persistent memory for repeat workflows
Monitor priorities and open commitments
Run scheduled tasks and deliver analysis
Pricing: Free and paid
Answer Calls With AI
Via IIEleven Agents
Reception.ai is an AI-powered receptionist for small businesses built on ElevenLabs’ ElevenAgents platform. It answers calls around the clock, books appointments in real time, and routes urgent requests based on business rules.
How you can use it
Answer business calls 24/7
Direct urgent calls based on custom rules
Support customers across 70+ languages
Handle after-hours and multi-location coverage
Pricing: Paid, free trial available
Jobs, announcements, and big ideas
Anthropic ships Claude Opus 5.5, a faster agentic coding leader at 40% lower cost.
OpenAI adds to its GPT-6 lineup with budget-friendly Sol and Luna models.
OpenAI calls for US-led global standards to govern autonomous AI research.
Rabbit debuts OS3, an agentic operating system with multi-device control and universal AI skill support.
Trump orders "artificial intelligence" renamed "super intelligence" across US government documents.
Xiaomi releases MiMo-V2.6, a multimodal AI series built on an open development approach.
Alibaba open-sources Qwen-Image-2.1, unifying image generation, editing, and transparency in one model.
Is this the future of robotics? I explain one very strange robot ⤵️
That’s a wrap! See you this Friday for more.
