Earning the Right to Spend on AI
JPMorgan's Apple Card • Microsoft's OpenAI Agreement • Meta's Thinking Machine Lab Attack • Meta's Apple Intelligence Attack • Meta's All-Star Pirates • Prototype iPhone 17 Pro
Feels like Meta's earnings call today after market close should be an interesting one given everything going on with the build up of the new 'Superintelligence Lab' and all the associated spend. Meanwhile, most analysts are expecting Meta's core business – the straw that stirs the AI cocktail, as it were – to show a slowdown in growth. So Zuck is going to have to be pretty good with the narrative here...
• Written on an M4 MacBook Air 💻
• Sent from London, England 🏴 Spyglass
🦙 Meta's 'MoE' Mistake
More details on how Llama's open source AI backfired...
🎞️ IMAX is More than Just a Bigger Screen, You Morons
Movie theater chains aim to slaughter their last remaining cash cow
Notebook
💳 JPMorgan Chase Set to Take Over Apple Card
At last, Apple's long national credit card nightmare is nearly over after a couple years of uncertainty. (Though it's still wild it has taken about a year since the JPMorgan talks were first reported.) On the surface it seems odd that Goldman wants to give up this business, but this report gets at why (beyond Goldman wanting to get out of the card business in general): Apple's insistence on a lack of late fees mixed with higher than usual default rates. And that's tied to fact that the card has a large potion of "subprime" customers – which is something you might not expect given Apple's adjacency to a more affluent consumer base. The mismatch here seems related to whomever was approving these cards? The whole thing has seemed rather mismanaged from the start. Sadly, we're likely also saying goodbye to Mastercard here (I don't really care about the card network but Apple Card was just my one Mastercard in a sea of Visa and AMEX). [WSJ 🔒]
🤝 Microsoft Nears New OpenAI Agreement
In other "nightmare is nearly over" news, this reports suggests we're just "weeks" away from a new agreement between tech's current favorite frenemies. And it reads like Microsoft being allowed to continue to use (and sell) OpenAI's technology even after AGI is declared – which itself suggests that could be happening soon. At the very least, well before the 2030 timeframe under which the two side's current agreement expires. Might this timing have something to do with the impending launch of GPT-5, or is it simply a coincidence? Regardless, this agreement – if it's consummated, which we simply can't assume until it's done at this point with these two – seemingly sets the stage for OpenAI to also shift to a PBC, which means actual equity stakes for folks such as Microsoft. This reports puts that in the "low- to mid-30% range" – which is exactly where I guessed it would net out nearly a year ago at this point. [Bloomberg 🔒]
🧠 Meta Fails to Invade Thinking Machine Lab
Some wild reporting in here by Kylie Robison (for her new newsletter) about Zuck's attempts to lure away a huge portion of Mira Murati's recently formed and funded AI startup. Not only did he apparently go after "more than a dozen people" at the 50-person startup, he offered one person over a billion dollars to make the jump to help steer Meta's 'Superintelligence' ship. Assuming that person was not Murati, who could that have been? One obvious guess would be any of her co-founders, who also jumped over from OpenAI – that would follow Zuck's playbook with another former-OpenAI executive's startup: Safe Superintelligence. After Ilya Sutskever turned down his 'Godfather' offer, Zuck poached his co-founder (and CEO), Daniel Gross. Murati has also reportedly already turned down Zuck's approach to buy the entire company. But unlike with SSI, "So far at Thinking Machines Lab, not a single person has taken the offer." Ouch. [Wired]
💸 Meta Does Not Fail to Invade Apple Intelligence
On the topic of Zuck's talent raids, a fourth person has apparently jumped from Apple's foundation models group, following the leader of that team, Ruoming Pang, accepting the 'Godfather' offer a couple weeks ago. While the overall number may be small within a massive company like Apple, this is clearly a major problem for their go-forward AI strategy and execution. And while Mark Gurman is reporting that they're "marginally increasing the pay" amidst these talent raids, Apple is clearly going to be far below market when the market has been raised to millions, if not tens of millions, if not hundreds of millions – if not billions – by Zuck. But beyond the money, which clearly doesn't always work by itself (see: above), Apple's problem is also clearly one of direction and the feeling that they're behind in AI with a roadmap that must pale in comparison to other places right now. Forget making a deal for any sort of tech, they're at the point where they need to make one simply for talent. [Bloomberg 🔒]
I Quote…
"You can collect a team of all-stars but that doesn’t mean it is an all-star team."
– A "senior person" at a rival AI lab talking about Meta's efforts to poach the best talent to build up their new 'Superintelligence Lab'.
That was my general stance when trying to think through if Mark Zuckerberg's strategy would actually work.
The same report notes that Zuck is hoping to "create a start-up-like unit within Meta focused on developing advanced AI technology that is unencumbered by the bureaucracy of the $1.8tn company." As I wrote when the initial Scale AI deal was still just rumored:
This almost sounds as if Zuckerberg is Steve Jobs in the early 1980s creating the Macintosh team within Apple. Such a comparison would flatter anyone, but it does read like that's essentially what Zuckerberg is trying to create here after Meta's initial attempts at AI haven't really panned out. Are there going to be pirate flags?
But those people, of course, were largely internal at Apple – as are seemingly most of the people joining the new Superintelligence Labs at Meta – the leadership, in a way, is more like the Traitorous Eight.
Loose Leaf
Anthropic's new valuation looks to be landing closer to $170B, which again makes more sense relative to the market (which makes little sense relative to anything else), though if xAI actually gets their $200B ask, that seems a bit wild. [FT 🔒]
Meanwhile, Cohere is closing in on a round that will push it well north of $6B. They're targeting $200M ARR by year end (for context, Anthropic is at $4B ARR). Much like with Mistral in Europe, it looks like Canada is rallying to boost Cohere's business prospects... [Information 🔒]
A new "Study Mode" for ChatGPT focuses on the Socratic method to make the service not just an "answer machine" for students. But will it work when the "answer machine" is just a click away? [Wired]
US District Judge Yvonne Gonzalez Rogers, last seen not taking any shit from Apple on the App Store nonsense, is now busy not taking any shit from Elon Musk and OpenAI/Sam Altman in the former's racketeering lawsuit against the latter. If this goes to trial next year, get the popcorn ready. [BI 🔒]
While Intel can't catch a fab break, Samsung just did in the form of a huge AI chip deal with Tesla. The 'AI6' chips would seemingly be meant for the Cybercab – and potentially Optimus and other robotic things. And these chips would be made in Texas, with Elon himself helping to oversee production. [Bloomberg 🔒]
Google shoving yet more AI into Chrome with generated store reviews. Seems like a small, rather crufty thing to add front-and-center? [TechCrunch]
One of my favorite SNES games as a kid, Mario Paint, is now on the Switch (Online). And yes, it works with the Switch 2's mouse functionality! [Verge]
As much fun as it might be to see the government trying to send IBM a bill for like a trillion dollars, I have a hard time believing that the US will be able to move to a percentage fee for patents. Lutnick! [WSJ 🔒]
I Spy...
Yes, it looks like someone was spotted testing what is very likely an iPhone 17 Pro prototype in the wild, which is fun, but hard to see why it's a bad thing for that employee – he's seemingly doing exactly what he's supposed to be doing. That is, putting the new device through some "real world" testing in the wild, while being cautious to make sure the device is cloaked in the "fat pack" so that no one can see what it actually looks like (aside from the camera system – which is why people know that it's likely the new iPhone).
There was a time when another Apple employee wasn't quite so cautious – and no, I don't mean the one who left it in the bar (or at home while his roommate seemingly snatched it). To this Apple employee I say, nice job. And nice shades.
