Good morning, {{ first_name | AI enthusiasts }}. Washington spent June squeezing the AI labs, with Anthropic's Mythos and Fable pulled and GPT-5.6's release reined in. Sam Altman's response arrived in the FT: invite the government in, officially.
The invitation starts with rules, via a US-led forum with real authority to regulate the industry. But it might end with equity, with OpenAI floating a 5% stake for the government to give the public a real cut of the upside.
In today’s AI rundown:
Altman pitches US-led AI safety forum, government stake
Rowan’s Corner: Are smart glasses going to replace phones?
Delegate team tasks to Claude inside Slack
TML, Bridgewater show the power of specialized AI
4 new AI tools, community workflows, and more
LATEST DEVELOPMENTS
OPENAI
Image source: Images 2.0 / The Rundown
The Rundown: Sam Altman just used an FT op-ed to call for a U.S.-led forum that would set AI safety standards and decide who can use the most advanced models — landing just as another FT report said OAI discussed giving the U.S. a 5% stake.
The details:
The op-ed came out of June's G7 summit in France, where Altman says he and other AI executives sat with heads of government to talk through AI regulation.
Altman pointed to the IAEA during the Cold War to police atomic energy, plus aviation and banking rules, as proof of concept for an international referee.
OAI also reportedly floated a 5% government stake in the company, also pushing for other U.S. labs to pay into a dividend fund to redistribute wealth.
Altman said “democratic institutions must not cede their responsibilities to AI labs“, and that “citizens and their elected representatives must make the rules.”
Why it matters: Both the equity and regulation discussions are gaining steam, with the latter growing more important in light of the Mythos saga. But as former White House AI advisor Dean W. Ball said, the question is whether the AI wealth is given directly to households, or to a government that may or may not deliver on its promises.
TOGETHER WITH FIN
The Rundown: Fin is the AI support agent leading teams are deploying today, resolving an average of 76% of customer conversations while cutting costs and complexity. Now available through AWS Marketplace, Fin lets you buy with existing cloud commitments and accelerate procurement.
Join the July 9 session, which covers:
How to buy Fin through AWS Marketplace and apply spend to your cloud budget
Why leading teams run Fin on enterprise-grade AWS infrastructure
How the Fin + AWS partnership drives faster resolutions and lower costs
What AI-powered support actually looks like running in production
ROWAN’S CORNER
Image source: The Rundown
Rowan: Mark Zuckerberg has been calling smart glasses the next platform for years. When I interviewed him at Meta Connect, I asked if he believes smart glasses could ever be good enough to replace your phone.
Fast forward to now: it’s estimated that Meta sells 8 out of every 10 smart glasses worldwide. With billions of people wearing glasses or contacts for vision correction, the path to replacing the phone looks less far-fetched than it did last year.
On top of that, curbing screen time is one of the biggest trends right now. Most people reach for a smartwatch to do it. Zuck is betting the real answer is on your face.
I wore the new Meta Glasses for the last week. Here’s my cold-cut review:
GPS, live translation, music, voice notes, and a camera built into the frames are genuinely great. Glasses beat the phone here easily.
Other delightful surprise features include “what am I looking at” for instant sightseeing breakdowns, and hands-free recipes while cooking.
But I'm still walking out the door with my phone. Battery is the first wall. With the display, audio, and AI running, heavy use drains it in a couple of hours.
The bigger hurdle is that the glasses only surface Meta's own apps -- so WhatsApp, Instagram, and texts show up, but the rest of my phone stays dark.
So no, they aren’t replacing your phone just yet. But Meta owns ~80% of the market, and with Ray-Ban, Oakley, and now Kylie Jenner on board, the shift is already underway.
AI TRAINING
The Rundown: In this guide, you will learn how to set up and install Claude Tag in Slack. This lets you and your teammates tag @Claude like a real teammate, then have Claude work in the cloud and report back when it is done.
Step-by-step:
Get Claude Team/Enterprise and Slack admin permissions. Then, in Slack, open Admin > Apps and Workflows, search for Claude, and install it
Find Claude under Apps, open it, click Home, and connect your account. Then open Claude Tag settings on the web and connect the tools it should use
Run @Claude connect in Slack, copy the pairing code into Claude Tag’s browser setup, choose the channel scope, add usage credits, and launch
Open Advanced in Claude Tag settings and switch from the expensive default model to Sonnet. Then run a test prompt in a dedicated Slack thread
Pro tip: After the first run, check the admin panel for token spend, plugins used, and memory files created. You can edit or delete persistent Claude Tag memory files there.
PRESENTED BY KOPIN
The Rundown: AI models are scaling faster than the infrastructure that connects them. Kopin and Fabric.AI are developing Neural I/o™, a new optical interconnect architecture that transforms programmable MicroLED pixels into ultra-fast optical transceivers, moving data with lower power and opening a new path beyond traditional electrical interconnects.
The Impact:
Eliminates electrical communication bottlenecks
Enables higher bandwidth with less power
Unlocks a new foundation for AI infrastructure
A leading technology in an estimated $100B industry
THINKING MACHINES
Image source: Thinking Machines
The Rundown: Mira Murati's Thinking Machines Lab and Bridgewater published research testing top models on news filtering investment tasks, with a small custom AI trained on the fund's judgment outscoring every frontier model while costing far less.
The details:
GPT, Claude, and Gemini variants averaged ~50% accuracy on six tasks, which included flagging emails, headlines, and reports of importance for an analyst.
Prompts written by the fund's own investors lifted scores into the mid-70s, still shy of the 80% mark Bridgewater says analysts need to trust a tool daily.
Training the open Qwen3-235B model on expert-graded examples via TML's Tinker platform resulted in an 84.7% on the fund's tests, at 13.8x less cost.
Murati framed the project as "experts improving AI that empowers experts," with Bridgewater planning models for more specialized tasks across the firm.
Why it matters: The assumption used to be that the frontier would steamroll smaller, specialized models, but TML and Bridgewater's numbers show how powerful (and cost-effective) those solutions can be. Companies don't always need a model that does everything, just one that is the best at their highly specialized work.
QUICK HITS
🎥 Seedance 2.0 - ByteDance’s frontier AI video model, now generally available
📱 Cursor for iOS - Cursor's new mobile experience for coding on-the-go
👋 Claude Tag - Tag @Claude as a shared Slack teammate to delegate tasks
⚙️ ZCode - Z AI’s agentic coding environment tuned for GLM-5.2
Palantir CEO Alex Karp criticized frontier AI giants on CNBC, saying enterprises “are paying for tokens that create no value” and the labs “are stealing weights and alpha.”
Microsoft launched Frontier Company, investing $2.5B on a plan to put 6,000 in-house engineers and sector specialists at client sites to help build and run AI systems.
China's Kling AI secured $2B in funding, with the Kuaishou video spinoff moving to push its global expansion after OpenAI shut down rival Sora.
The Wall Street Journal reported that SpaceX demoed an xAI-powered phone prototype for IPO investors, with Elon Musk dismissing the claim as “utterly false”.
Anthropic reportedly approached Samsung about manufacturing its own AI chip, coming after poaching Clive Chan from the OAI team behind its Jalapeño chip.
COMMUNITY
Every newsletter, we showcase how a reader is using AI to work smarter, save time, or make life easier.
Today’s workflow comes from reader Anonymous:
“I'm 67 years old and have been using ChatGPT personally and professionally for almost 2 years. When I first started, I worked in social media in the beauty industry. I've since walked away from that and become an author of children's bedtime stories. ChatGPT has helped me learn how to write, illustrate, and publish on Amazon KDP — I have two books published and am almost finished with my third.
I started this project in March, and I knew nothing. I now use ChatGPT, Leonardo AI, and Topaz Gigapixel. I'm setting up the systems and foundation to build a whole world of merchandise, toys, and a brand. I would not have been able to even start without AI.”
How do you use AI? Tell us here.
Read our last AI newsletter: Anthropic’s Fable returns worldwide
Read our last Tech newsletter: WhatsApp’s ‘usernames’ reservation race
Read our last Robotics newsletter: Apptronik opens massive ‘Robot Park’
Today’s AI tool guide: Delegate team tasks to Claude inside Slack
RSVP to next workshop on July 8: Create Short Form Videos with AI
That's it for today!
See you soon,
Rowan, Zach, Shubham, and Jennifer — the humans behind The Rundown
