I’m back from vacation. Did you miss me? The AI news cycle certainly did not pause while I was trying to unplug in Greece. Oh well.

The first time I interviewed Tom Alison, about a decade ago, he ran engineering for News Feed and drew me a diagram on a whiteboard to explain how ranking worked. He runs all of Facebook at Meta now. When we sat down in Los Angeles on Friday, I asked where AI ends up in what employees still call the blue app.

“We talk about this all the time,” he said. “Should Facebook be a human-only network forever or not? And we haven’t made a strong stance on that.”

I haven’t heard anyone at Meta say that publicly, and it’s telling that Allison left it open in the same week Facebook began rolling out a free verification badge built to prove a person is real. Mark Zuckerberg has spent the past couple of years describing AI-generated content as a third era for social media, predicting that AIs and humans will co-exist in feeds together. Alison’s version is looser.

Read more

Schedule note: I’m taking the next two weeks off for some much-needed R&R. I trust (not) that no major AI news will break while I’m gone.

Anyone closely following OpenAI’s palace intrigue wasn’t surprised when Fidji Simo announced yesterday that she isn’t returning to her job as the company’s number two. Greg Brockman absorbed most of her remit in mid-May on a non-interim basis, all but confirming that she wasn’t coming back. Simo is staying on as an advisor while she continues to work on her health, and I’m told she has talked with Brockman almost daily throughout her medical leave.

Simo originally hoped that her leave would last a month or two at most, according to my sources. The AI race doesn’t idle, though, and while some companies can maintain velocity on autopilot when a senior leader steps away, OpenAI clearly couldn’t.

Read more

When I got Alexandr Wang on the phone ahead of Meta’s Muse Spark 1.1 release, I wanted to know how serious Meta was about its new API business for selling access to the model, which now puts it in more compeitition with OpenAI, Anthropic, and Google. Wang said that Meta sees the API as a real business, not just a learning exercise. Tokens are “gigantic and growing very, very quickly,” he told me, and capturing even a slice of that “can be a very meaningful business even for Meta at our scale.”

Every new API account comes with $20 in free credits, then $1.25 per million input tokens and $4.25 per million output, making it “one of the most attractive pricing levels in the market,” as Wang put it. The API also plugs into the OpenAI SDK, so developers can point their existing stack at Meta’s model. Wang said that was deliberate: “We wanted adoption of our model to be very efficient and easy for developers starting out.” A Meta-native stack will come with time. Claude SDK compatibility isn’t part of this release, but he didn’t rule it out either.

Read more

“Over time, we think this will also unlock the ability to use voice as a kind of primary interface to computing, and to manage increasingly complex long-running agentic work.”

ChatGPT Voice product lead Atty Eleti said that near the end of OpenAI’s press call this morning announcing GPT-Live, its new generation of voice models. When I asked at the very end of the Q&A whether the models were devised for Jony Ive’s forthcoming AI device, OpenAI comms, naturally, shut me down.

A voice model that’s smart and fast enough to be the primary interface for computing is exactly what OpenAI needs for the device it’s building with Ive. Sources tell me the first device is likely to be a sensor-laden, ambiently aware, puck-like object that you control with your voice and that proactively engages with you throughout the day.

Read more

The holy grail for the frontier labs is a super app that combines the power of AI coding with the simplicity of a chatbot, and I’ve spent time reporting on how OpenAI and Anthropic are each chasing it.

Today Anthropic inched closer. Claude Cowork, or Claude Code for people who don’t code, is being merged with the Claude chatbot interface on desktop, web, and mobile for Max users, and Cowork sessions can now keep running in the cloud after you close your laptop. Moving Cowork sessions to the cloud is the real step forward, because scheduled tasks can now run without a desktop machine powered on or a laptop propped open.

Mike Krieger, who ran product for all of Anthropic before recently moving to its shunkworks Labs group, sketched this exact step for me in May:

“Some of these products naturally suggest their next evolution. I’m a huge Dispatch power user. Dispatch lets you access your Cowork remotely, but you gotta run your computer. So the next logical thing is, well, this would be great if I didn’t have to leave my computer on all the time. Claude Code has been on that journey… Claude Code Remote I use all the time. I’m out on the go, I can kick off the coding task, and by the time I’m back it’s often put up a pull request. You can imagine similar things for Cowork.”

Cowork threads still don’t fully sync across desktop and phone. What could really move the needle on non-early adopters getting access to a multi-agent harness like Cowork is when it’s finally merged with the chat interface on the Claude mobile app, which may be what Krieger hinted at on X today: “Keep an eye out for even better Chat + Cowork integration soon, too.”

Cat Wu, who leads product for Claude Code and Cowork, told me in March why Cowork beats a normal chatbot.

Read more

It’s Wednesday evening, but doesn’t it feel like Friday? Happy July 4th weekend. This will be my last send for the week.

While I was on a call earlier today, I got the push alert from The Wall Street Journal: “[Elon] Musk has developed a prototype for a handset-like device designed to reshape how humans interact with AI. SpaceX recently showed it to investors.” For months, I’ve heard rumors of some kind of SpaceX handset in development, though, as the Journal noted, it’s in the early stages, and plans for it could change.

Read more

It’s Tuesday night on the West Coast, and Fable cometh. Will Anthropic getting clearance to re-release it prompt OpenAI to push out GPT-5.6 this week, even though its employees are supposed to be on vacation? We shall see!

Today I’ve got a special report from my day at a unique gathering of AI researchers, CEOs, and academics in San Francisco, where the vibe was one of overwhelming concern about the concentration of power at the frontier labs.

I got to the Exploratorium in San Francisco on Tuesday morning for Open Frontier, a one-day “working meeting” of AI researchers and industry folks convened by Andy Konwinski’s Laude Institute. Conversations throughout the day centered on concern about a growing concentration of power in AI, with too much talent, compute, and data pooling inside a few closed labs.

Konwinski’s opening framing was that open AI research is critical not only to the health of the field but to democracy more broadly, and that the frontier is now closing, with AI’s control increasingly being “pulled back behind closed doors.” His charge to the room: the frontier “will not close on its own. It closes through a series of decisions” made by the people in it. “We are not spectators.”

China and closed labs are winning

Read more

Even though the OOO auto-replies are on for a lot of folks, I’m here. On Tuesday, I’m headed to Laude’s Open Frontier forum in SF, where a bunch of leading AI researchers will be sharing thoughts about the state of open source, concentration of power, and where things are headed next. Stay tuned for takeaways. Today, I’ve got an inside look at Cursor’s big launch and a feed check.

The recent rise of AI coding agents has led engineers to use computers in unnatural ways. “You see engineers walking around with little devices to keep the laptop from closing, to keep their agents alive,” Kevin Niparko, Cursor’s product lead, told me today.

The new Cursor iPhone app hopes to change that. Released today on the iOS App Store, it lets you control cloud agents or remotely control the Mac minis on your desk, with integrated voice control, push notifications, and live activity tracking.

The new app is part of a growing trend of AI coding moving to phones. Codex has lived inside ChatGPT since May, and the Claude mobile app has been able to interface with Claude Code on a connected desktop for longer. Cursor’s bet is that a dedicated app beats one bolted onto a chatbot.

Read more

Happy Friday. I’m en route back to Los Angeles after a busy Cannes Lions week of meetings and interviews. Today, I’ve got an interview with one of Adobe’s top execs from Cannes, and a feed check on OpenAI’s big model release today, good links, and more.

On the punishingly hot final day of Cannes Lions, I sat down with David Wadhwani, who runs most of Adobe’s businesses, on the company’s stretch of sand behind the Palais. I wanted to hear his reaction to the argument that has gotten loud in Silicon Valley and on Wall Street: that AI is chipping away at Adobe at every layer of its business.

Mad Men to AI

This year was Adobe’s largest Cannes presence yet, with a creator-themed beach space, podcast studios, and a bunch of panels, including one with Wadhwani and podcaster Steven Bartlett.

Wadhwani has come three years in a row. This was my first time, and we traded notes on how the festival has changed over the years, from the Mad Men, agency-led advertising era, to when tech platforms like Google and Spotify became dominant, and how this year, AI and creators dominated the week.

“Demand for brands using creators has gone through the roof,” he told me. That much was evidenced this week at Cannes: UTA built its own beach space for the first time, OpenAI and Anthropic both sponsored creators who attended the festival, Substack and Beehiiv sent bigger delegations than ever, and there was, of course, my creator economy panel at Yahoo’s beach space on Monday. For its part, Adobe sponsored a new, official creator track that brought a bunch of creators to participate in panels and mingle with CMOs all week.

Token pricing

In Silicon Valley, the story on Adobe is death by a thousand startups, combined with competition from more nimble, younger rivals like Canva and Figma. Nearly every week, I see a new AI startup coming for part of the Adobe stack, each promising to one-shot what used to take a Creative Cloud subscription to get done.

Read more

On Monday evening, to kick off Cannes Lions, I hosted two conversations at Yahoo’s beach stage: a conversation with CEO Jim Lanzone, followed by a creator-economy panel with Tyler Denk, co-founder and CEO of beehiiv, Casey Lewis, author of After School, and Cat Goetze, founder of the AI media brand CatGPT. Everything we discussed, naturally, centered on AI.

My main takeaway from both conversations was that, despite all the fear about what AI will do to jobs, everyone agrees that the value of unique, original content is rising.

No content, no answers

I asked Lanzone why the world needs another chatbot when it already has ChatGPT and Claude. His pitch for Scout, Yahoo’s AI answer engine that launched in January and remains in beta, is that every large language model is “created the same way” by “suck[ing] down the entire internet.” Yahoo’s edge, in his telling, is its decades of search query data, direct relationships with 700 million monthly users, and work with publishers across its news and finance verticals. He said the company is“applying AI search technology to that data” to produce “completely unique answers.”

He thinks the AI industry got the social contract with publishers wrong by hoovering up the web without sending much of anything back, and that licensing payments from AI companies will never scale enough to keep the internet alive. His proposed fix with Scout is the oldest move in search: prominent links that route traffic back to the people who did the reporting. He knows the limits of the alternative. Only so many publishers can sell subscriptions, he noted, with a glance my way: “You’re one of them.”

My last question to him was about the state of the news business, given his current role at Yahoo and his tenure running CBS’s digital business a decade ago. Lanzone pushed back on the popular idea that trust in media is broadly collapsing, arguing the erosion is concentrated in one category, politics and current events, and not sports, finance, or the movie and restaurant reviews people read every day. He thinks more people are producing news now than ever, just outside the gatekeepers who used to control it, and that real journalism "with real integrity" will always have a market and its defenders. As an aggregator sitting atop hundreds of those sources, he said Yahoo is "pretty bullish."

What creators shouldn’t give to AI

All three of my creator economy panelists use AI constantly, and all three protect the same thing: their own ideas.

• Goetze: “Never let it come up with your ideas.” She uses AI for what used to be impossible, like work that “would have taken a team of 10,” not just to move faster. Her Bluetooth phone brand did six figures in revenue in just a few days, and she’s opening Cat Labs, a creator-first product studio in LA, built for other people who build things together in a room.

• Lewis: She runs every draft of her newsletter through ChatGPT, Claude, and Yahoo to catch typos, but won’t let AI write or generate ideas. “They will never replace thoughtful editors.” She leans on quotes from articles in her newsletter because journalism is “coming from the source’s mouth.”

• Denk: Beehiiv’s recently launched MCP server has already surpassed 100 million requests, and the case for it is augmentation: put AI on the admin side so creators spend more time creating, not less. He said the best creators “bring you perspective and experience that can’t be replicated by AI.”

How the kids feel about AI

Read more

Hello from Cannes Lions, where the humidity here could suffocate even the thirstiest marketer. This week, I’ll be bringing you what I’m hearing at the axis of AI, creators, and advertising, starting today with OpenAI’s industry pitch for ads in ChatGPT. Also, due to the time zone change, I’ll be sending issues during morning Pacific time through Friday.

Monday morning, in a villa off the Croisette, OpenAI sat press and analysts under umbrellas and misters in the heat for its first-ever briefing at Cannes Lions, the ad industry’s yearly blowout. OpenAI sent roughly 60 people this year. Four years ago, its pitch to this crowd was Mira Murati onstage in the Palais talking about AI and creativity, when all OpenAI had to show marketers was DALL·E. This year, it came to sell ads in ChatGPT.

Read more

Andy Konwinski has a phrase for where the AI industry is headed if nothing changes: “feudalism with better branding.”

Two or three companies end up owning the foundational resource, while everyone else rents access on their terms. Konwinski co-founded Databricks and Perplexity, helped build Apache Spark, and is now spending $100 million of his own money through the Laude Institute he co-founded to keep that future from coming true.

The problem Konwinski and his colleagues at Laude have landed on is the hollowing out of academia by the frontier labs and Big Tech, which are publishing less and less research publicly. Pay gaps between higher ed and the labs have widened from roughly 2-3x a decade ago to over 10x today. The highest payers also increasingly control the frontier compute that researchers need to do the work.

On this week’s episode of ACCESS, Konwinski says his worry isn’t safety so much as who gets to define the future of AI. “There’s a difference between safety and centralized control,” he says. “The threat to democracy of centralized control is real.”

Listen or watch wherever you get podcasts.

Laude turns one year old next week, and Konwinski has moved fast. The institute has handed “Slingshot” grants to 15 research teams (one built a benchmark Dario Amodei name-checked for the Claude 4 launch) and is backing “Moonshot” labs run by people like Stanford’s Erik Brynjolfsson and Terence Tao’s group at UCLA, with a board that includes Google’s Jeff Dean and Cohere’s Joelle Pineau.

From where I sit, few people command as much respect across the research community as Konwinski. Earlier this week, Ellis Hamburger and I sat down with him in Pasadena for the podcast. We also discuss:

• The AI community’s backlash to Fable. When Anthropic shipped its new Fable model, it quietly downgraded certain kinds of use, including cyber and bio questions and anything resembling AI research or distillation, to Opus without disclosing it. Export controls then locked out foreign nationals, including Anthropic’s own (and Laude-affiliated) Andrej Karpathy. Konwinski argues that the bigger issue wasn’t just the policy itself, but that a single company felt it had the authority to decide who could access a foundational capability.

• His answer isn’t to outspend the labs. It’s to build an open ecosystem that fights differently. “They fight in roads, we can fight in ditches,” he says, betting that scarcity breeds innovation and that an open ecosystem can own the benchmarks and context layers that decide what models optimize for.

• He sees a leadership crisis in AI. Alongside the consolidation of power, Konwinski's biggest worry is what he calls "an absence of leadership.” The incentives for CEOs and VCs are “completely wrong," he argues. Part of that vacuum is the researchers themselves. "AI researchers have left society behind entirely," he says, and are largely out of touch with how people feel when data centers go up in their backyards.

A MESSAGE FROM MY SPONSOR

Paid training and a job.

The AI revolution will bring historic opportunity. America’s Workforce Academy, built by Meta, is helping train those who will build it.

The program offers paid training in electrical, plumbing, fiber installation and many more, with a job upon completion.

The future is for everyone.

Learn more.

Feed check

Read more

Read more

For those curious, my about page explains how ads in the newsletter work, and the line that’s drawn between them and editorial.

The stakes going into Tuesday were about as high as they get for Snap.

Evan Spiegel has spent more than a decade and billions of dollars chasing augmented reality glasses. He’s called this year’s Specs glasses launch a crucible moment, and he’s making the bet during the roughest stretch of the company’s life as a public company.

At the Augmented World Expo (AWE) in Long Beach, the AR industry’s biggest annual gathering, Spiegel partly unveiled a consumer version of Specs and opened pre-orders for the $2,195 device. By Tuesday’s stock market close, Snap’s stock had fallen almost 10%.

The fully standalone AR glasses Spiegel showed represent the same vision he’s chased for a decade. But the rest of the industry has moved in the opposite direction, and that bet is the one that's working. Meta and soon Google are shipping lightweight, normal-looking glasses built around cameras, speakers, and AI-powered voice assistants.

Spiegel waved that approach off to me when we spoke on Monday, calling it “a phone accessory.” Onstage at AWE, he took a more direct swing that, depending on how you look at the situation, could also be seen as a self-own: “Those copycats up north aren’t going to be stealing this one.”

Read more

A reminder that I’ll be at Cannes Lions next week for a Monday evening of programming with Yahoo, plus a Tuesday panel called “Attention, AI and the Future of Marketing” at the OpenAP Villa.

The day before the SpaceX IPO, I called Shaun Maguire, the Sequoia partner who led the firm’s investments in the company.

Maguire spent 17 years in LA’s space scene and earned a PhD in physics from Caltech while studying black holes. He watched Friday’s IPO from Starbase with Musk. Sequoia’s roughly $2 billion investment in SpaceX is worth more than $20 billion, its biggest dollar return ever.

Much has been written about how the SpaceX IPO will change the financial lives of employees and investors. I wanted to know what Macquire thought it meant for the rest of the tech industry. With new startups, “there’s this reorientation happening right now from software to a lot more hardware,” he says.

Read more

Read more

In addition to my Monday evening of programming with Yahoo, I’m excited to announce that I’ll be participating in a panel titled “Attention, AI and the Future of Marketing” with Joe Marchese of Human Ventures and Bryson Gordon, Microsoft’s CVP of global marketing, at the OpenAP Villa during Cannes Lions. Let me know if you want to attend.

Read more

I’ve been in Cupertino all day at Apple Park for WWDC. My takeaways are below. Keep reading for a scoop on Anthropic's next big model release.

Right after Apple’s WWDC keynote finished, I joined dozens of other journalists, influencers, and analysts in a dimly lit theater at Apple Park to hear an AI 101 class and Q&A with Craig Federighi and other Apple execs. Tim Cook and incoming CEO John Ternus sat in the front row.

Apple has never done an on-the-record press Q&A like this after a big keynote. There was no live broadcasting, but below are my notes and takeaways:

Read more