Today’s theme is an old one. You can’t have your cake and eat it too.

And there is a wrinkle in that phrase worth knowing. The older version, the one that actually makes sense, runs the other way around. You can’t eat your cake and have it too. Once it’s eaten, it’s gone. That’s the whole point.

Today is about three sets of people trying to do exactly that. Eat it, and still have it.

Anthropic and Dario Amodei, on open weights.

OpenAI and Anthropic together, as arch-rivals quietly on the same side in Washington.

And AI investors, deciding this week whether they are terrified of AI spending or embracing it. There is evidence of both.

I have been on these topical threads for a while now. Friday it was threading needles. Yesterday it was Nvidia assembling an alliance. They are one long tussle, and today they edged forward in different ways.

This is the AI Tech Wave at the point where the professing and the doing are different sides of the same coin. Three events, each with my Take, then my Overall Take. Plus a Gadget AI on Apple’s home push, and two questions. Let’s get started.

(1) Anthropic and Dario Amodei on Open Weights

MP TAKE: Dario Amodei took to writing yet another essay. (links below) This one is called “Our position on open-weights models,” and it is Anthropic talking on the record about this whole open source AI issue.

They were the ones very prominent in their absence from recent debates on the topic. The only major name that did not sign the big industry letter that Nvidia founder/CEO Jensen Huang drove over the weekend. Almost forty companies signed. Thirty-seven, to be precise. Including some unusual ones like OpenAI itself and Google, who are on both sides of this in many ways.

So this was Dario’s way of getting his say out. Basically that his company is not against open-weight models in general.

Note he stresses open weights, not open source.

Open source is the broader one. It lets you see how the thing was built, the training, a lot of the other data around it. Open weights are the numbers. They matter, but they are not everything. It is “open source light”, the way I think about it.

Here is the carve-out. He does not want a ban. He does want mandatory safety testing before release. Which his company already does a lot of, and has a head start on.

And to his credit, he says the quiet part out loud, right near the top:

“It would protect US AI companies from competition, but that has never been my goal.”

That is a candid acknowledgment. The policies he favors would advantage the incumbents against everyone coming from behind. Meta. Elon’s SpaceXAI. And most of the rest of the world, including China. He is telling you the effect while disclaiming the intent, and both of those things can be sincere at once.

So he is doing what a CEO should do. Especially in this knotty, gnarly world of global AI right now, where the CEOs are trying to be more like politicians than most politicians.

Which is also what the other CEO is doing. Jensen wrote the letter and started the Open Secure AI Alliance that I wrote about this morning in AI-RTZ #1161.

Two CEOs, two essays, one chess game. With global consequences.

Sources:
– Anthropic, ‘Our position on open-weights models’
– Axios, ‘Anthropic CEO Dario Amodei says he does not support open-weight AI ban’

For longtime readers:
– ‘Nvidia Assembles the AI Avengers’ in AI-RTZ #1161
– ‘Anthropic & OpenAI now against China open source AI’ in AI-RTZ #1156

(2) The Arch-Rivals, Working Together Behind the Scenes in Washington

MP TAKE: Both of them want similar things out of Washington for a change. A policy that adds regulation. That slows their rivals, here or in China. And that takes a more ‘anti-open-source AI’ stance than either seems willing to say on the record.

And remember, these companies are siblings.

Dario and Sam worked at the same company, closely, on the earlier versions of GPT. Amodei arrived at OpenAI the year after it was founded and rose to run research, straight through the years that produced GPT-2 and GPT-3. Then in 2021 he and his sister Daniela left, with a group of early OpenAI people, and built Anthropic across the street.

I wrote about that split at length in OpenAI & Anthropic, different birds of a feather, AI-RTZ #928. Same ancestry, starkly different business models. One went enterprise and disciplined. The other went consumer and globally ambitious. Amodei’s own description of Altman’s approach, in that piece, is that he “constitutionally just wants to YOLO things.” He meant living with the driver that ‘you only live once’.

Anthropic is the Pepsi that is now the Coke of the AI industry. That is the reversal nobody quite expected.

And the two companies today are arch-rivals to the point where back in February, both on stage with Prime Minister Modi in India, they would not actually hold hands for the group photo. That famous shot is in the links below.

All of that is on the back bench right now. In Washington they are allies of convenience. Machiavelli style. Keep your enemies close, and that sort of thing.

But they are also going to be very hard at work against each other. Both are trying to get mega AI IPOs done this year and next. And they are in a race for commercial customers, enterprise customers, and AI coding customers for their leading models.

They have come a long way, with a lot of twists and turns, to have their anti-open-source cake and eat it too.

In the market, rivals. In the hallways of DC, allies. For now.

Sources:
– The Information, ‘OpenAI and Anthropic are quietly teaming up in Washington DC’
– AP, ‘Modi’s AI Summit turns awkward as tech leaders Sam Altman and Dario Amodei dodge contact’

For longtime readers:
– ‘OpenAI & Anthropic different birds of a feather’ in AI-RTZ #928
– ‘Anthropic & OpenAI now against China open source AI’ in AI-RTZ #1156

(3) Investors, Fearful and Greedy at the Same Time

MP TAKE: That is what a market is all about. As Warren Buffett has put it, borrowing from Ben Graham, in the short run it is a voting machine. Long term a weighing one. Every once in a while you actually have to weigh things and take a side.

Today the investors are relatively schizophrenic. Look at the moves in memory chip stocks, which have seen dramatic volatility both here and around the world, and especially in South Korea.

And I want to touch on the Korea part, because Korea’s markets are extraordinarily volatile right now, for structural reasons.

Start with concentration. Two memory companies, Samsung and SK hynix, and I have written a lot about both. Each now carries a trillion-dollar-plus market cap. Together they are more than half of the KOSPI index. Over half the market capitalization of a G20 economy’s stock market, in two chip companies. There is no US equivalent, and it is not close.

Then add the people. Of South Korea’s fifty-plus million, over fourteen million, mostly young retail investors, have been meme-stock trading those two names and the index. Better than one in four of the entire population. And a lot of it through leveraged ETFs.

High risk, high reward. And when it goes very volatile, if you are trading on debt, you get margin calls.

Which is exactly what has been happening. Over a million people hit with margin calls in recent days. Circuit breakers tripped on the market. The leadership of the country has had to get involved.

That is not what is going on in the US of course. Our volatility is more contained. Two sides of the same coin, with the pendulum swinging a lot wider in Korea than here.

Meanwhile, on the corporate side in the US, investors are trying to get their heads around the continuation of the circular financing deals, especially led by Nvidia this week.

A $250 billion backstop to OpenAI’s spending. A big multi-billion-dollar investment in Ilya Sutskever’s Safe Superintelligence, a company with barely a product and no revenue. And billions financing data centers in Texas and elsewhere, running on Nvidia’s own chips.

All of it adds up to something on the order of $750 billion or more in announced AI commitments. The market is a little spooked by that. And it is notable that Nvidia’s own credit default swaps also hit a record.

That is a signal from the credit market, not the equity market.

And while all this is going on, the equity side is waiting for the quarterly results later this week. I wrote about this last week when Alphabet reported and raised capex to over $200 billion. That is what is anticipated this week from Microsoft, Meta, Amazon and Apple. Apple is obviously more contained on AI spending than the others.

What matters is less the quarter. It is the capex line, and how they frame the growth.

Sources:
– WSJ, ‘Which Tech Giant will Blink first on AI Spending’
– Axios, ‘Nvidia reignites “circular” AI concerns as it weighs OpenAI financing guarantee’
– CNBC, ‘US and Korean tech stocks are now tightly linked’

For longtime readers:
– ‘Latest report on an AI Bust and the Global Economy’ in AI-RTZ #1134
– ‘Google & Tech Cos Backing Up the Truck on AI Infrastructure’ in ARD #125
– ‘Fear & Greed Drive Risk Mitigation’ in ARD #92

MP OVERALL TAKE

So far I think we will be okay through this week and beyond.

On the open source machinations, the White House is likely to thread the needle here. They are getting ready for the big second meeting between the two Presidents at the White House, and AI is going to be very much front and center.

So there is a lot of gamesmanship going on around all of this. Chess, checkers, whatever you want to call it. And a lot of these are bargaining pieces. Open source AI models. Chips. Trade. Access to the latest chips. You had Moonshot again officially saying they could really use more Nvidia GPUs for their next generation models. There is a lot of horse trading to be done.

On the investor side, the fear and greed will see through this week’s earnings results. Investors should generally continue to lean in, given there are still good signs of demand for AI compute. We have not seen aberrations on that side of the equation, although we are all watching very closely.

Here is what I keep coming back to. Everybody today is trying to eat the cake and still have it. And for the moment, they are managing both.

The conditions right now allow two incompatible things to sit side by side.

So for now, it seems there is time to enjoy the cake while we still have it.

GADGET AI

Apple Readies Three to Five AI Gadgets

MP TAKE: Apple is getting ready with about three to five gadgets. Three of them as early as later this year, and a couple into next year.

They are all surrounding this category of home AI gadgets. Things that listen to what you say and do around the house. Speakers. There is a HomePod mini being talked about. There is an Apple TV product, the hardware streaming device, upgraded with better chips.

And there is a new product, potentially a HomePod with a small seven-inch screen that has Siri AI, which is Apple’s new ace card in AI.

Then supposedly two more devices for next year. One with an articulating bigger display that follows you around and remembers things. The second even larger, wall-mounted on magnetic clips. There is a very detailed piece in Bloomberg on all of it.

All of this arrives in the teeth of ‘RAMageddon’, the supply constraints and memory price spikes we have been seeing. And will see a lot more of over the next two or three years. Very challenging times for all of these companies.

But the broader point about these gadgets is not just Apple. It is their peers. Google. OpenAI, planning a bunch of gadgets with Jony Ive after that acquisition. Meta. Amazon, of course, very big in the space. And SpaceXAI, where Elon has talked about phone-like gadgets too.

Two of them want to do more than home devices. They want smartphones that are not actually smartphones, connected via satellite networks. That is SpaceXAI and Amazon. And OpenAI and Meta are considering smartphones as well, competing directly with Apple.

So a lot of choices coming to market over the next eighteen to twenty-four months.

Apple is different here, and I want to summarize it again.

Keep in mind Amazon is obviously a retail and advertising business. Google is an advertising business. Meta, of course, is an advertising business. And for all of them, a device in the home is a sensor. It is there to snarf up all the data.

And not just waiting for a wake word, the way these things have been designed to do for the last decade with Amazon Alexa and Echo, where they only respond to the question.

Now, with your permission, they will be watching and recording everything. And all that data gets snarfed up to train models and, hopefully, give you better AI results through all of it.

Apple is the only one that is going to adhere to the card of trust, safety and privacy. They have played it for years and they will probably take it a step up in this AI era.

That is going to be a differentiated card. And it remains to be seen how the broad mainstream market of hundreds of millions of people really responds to a world where these devices really do intrude on our lives.

That is supposedly a trade, where the bigger benefits are the learning from the community data and the results we get from it. That is the essence of the AI trade. Whether it is in devices, or our smartphones, or chatbots, or AI agents.

And that is what is now going to play out in home devices.

Sources: Bloomberg, ‘Apple set to make big smart home push with Siri AI at center’

For longtime readers: ‘Apple Intelligence and Siri AI with Google and Nvidia Inside’ in AI-RTZ #1113 · ‘RAMageddon’ really here to stay in AI-RTZ #1145

Q1. What is my most-used AI gadget at home?

Answer: The smartphone.

I have all the gizmos. The HomePods, the things with screens. But I use them mostly for timers, asking for the time and weather, and the occasional bit of music.

For everything else it is this, because it has all my data, all my applications, lots of memory, the connectivity. I still maintain that the smartphone form factor is the AI gadget we have been waiting for.

And the company at the lead on it is Apple, in my mind. With Google and Android close behind. But I think generally Apple over Google, because of exactly the difference I talked about. Their business model, versus us being the product for everything else.

Q2. What would change that?

Answer: I am intrigued by the upgraded Apple TV.

But for all of these devices that deal with the TV, the core feature I really want is essentially answering the question, constantly, especially with family around: “what to watch”.

That is a critical question we all struggle with. Right after the other most important question in our home, which is “what to eat”.

And these are typically discussions that take longer than the meal or the program.

It is the tyranny of choice, which I have written about in multiple essays here. And all of it is getting turbocharged with AI, especially with the abundance of content exploding with agents and everything else. It is going to go up a lot more.

WRAP

Today’s AI-RTZ #1161, on Nvidia assembling the AI ‘Avengers’, the Open Secure AI Alliance, that big collection of over thirty-seven companies, and it will be forty or more in the next few days. All of it around open-weight AI models from China and beyond. Anthropic, of course, regardless of their essays, is going to be battling that.

AI Ramblings Daily on AI-RTZ is here to think through AI and reset. Together.

Tomorrow, ARD 129 and AI-RTZ #1162.

Thanks for joining us today, AI Curious Folk. Stay tuned.

— MP

Full Source Reading

For the broader context, see the canonical sources for ARD 128, in today’s narrative order:

Event 1. Anthropic and Dario Amodei on Open Weights

• Anthropic, Our position on open-weights models

• Axios, Anthropic CEO Dario Amodei says he does not support open-weight AI ban

• The Information, Anthropic’s Claude Code Reigns despite rising interest in Codex, Open-Source Models

Event 2. OpenAI and Anthropic in Washington

• The Information, OpenAI and Anthropic are quietly teaming up in Washington DC

• AP, Modi’s AI Summit turns awkward as Sam Altman and Dario Amodei dodge contact

Event 3. Investors, Fear and Greed

• WSJ, Which Tech Giant will Blink first on AI Spending

• Axios, Nvidia reignites “circular” AI concerns as it weighs OpenAI financing guarantee

• The Information, Why Investors Shouldn’t Panic about Nvidia’s commitments

• CNBC, US and Korean tech stocks are now tightly linked

• Reuters, South Korea’s KOSPI slumps as chipmakers sell off

Gadget AI. Apple’s Home AI Push

• Bloomberg, Apple set to make big smart home push with Siri AI at center

Clips from today

Clip 1. Your AI Home Gadgets Are Sensors

Look at what these companies actually are underneath the hardware. Amazon is retail and advertising. Google is advertising. Meta is advertising.

MP Take: Apple is different here, and I want to summarize it again. For all of them, a device in the home is a sensor. It’s there to snarf up all the data, not just waiting for a wake word the way these things have been designed to do for the last decade with Amazon Alexa and Echo, where they only respond to the question. Now, with your permission, they’ll actually be watching and recording everything, and all that data gets snarfed up to train models, and then gives you hopefully better AI results through all of that.

Clip 2. Apple Is Readying 3 to 5 New AI Gadgets

Apple is getting ready with three to five gadgets. Three as early as later this year, a couple more into next year.

MP Take: They’re all surrounding this category of home AI gadgets. Things that listen to what you say and do around the house. Speakers, a HomePod mini being talked about, an Apple TV product upgraded with better chips. And a new product, potentially a HomePod with a small seven-inch screen that has Siri AI, which is Apple’s new ace card in AI. Then supposedly two more devices for next year. One with an articulating bigger display that follows you around and remembers things. The second even larger, wall-mounted on magnetic clips.

Clip 3. Apple’s Privacy Card, and the Real AI Trade

Apple is the only one of them adhering to the card of trust, safety and privacy. And that is going to be a differentiated card.

MP Take: They’ve played it for years and they will probably take it a step up in this AI era. It remains to be seen how the broad mainstream market of hundreds of millions of people really responds to a world where these devices really do intrude on our lives. That’s supposedly a trade, where the bigger benefits are the learning from the community data and the results we get from it. That’s the essence of the AI trade. Whether it’s in devices, or our smartphones, or chatbots, or AI agents. And that’s what’s now going to play out in home devices.

Clip 4. Nvidia’s $750B Circular AI Bet

The other thing investors are trying to get their heads around is the continuation of the circular financing deals, especially led by Nvidia this week.

MP Take: They just announced a $250 billion backstop to OpenAI spending. A big multi-billion-dollar investment in Ilya Sutskever’s Safe Superintelligence company, of OpenAI fame, one of the chief technologists there. Again, a young company, barely a product and no revenue. And they’re financing billions in data centers in Texas and other places using their own chips. All of this is adding up to something on the order of $750 billion or more in announced AI commitments. The market is a little bit spooked about that. And it’s notable that Nvidia’s own credit default swaps also hit a record. That’s a signal from the credit market.

About AI Ramblings Daily (ARD), and AI-RTZ

Both are daily. Both are free. Both are about AI. But they’re different mediums carrying different messages.

AI-RTZ is the morning text, a deeper written take on one idea, published by at least 5 AM EST. Today: post #1161.

AI Ramblings Daily is the afternoon video + podcast, my ad hoc takes and perspective on the day’s AI issues and news flow, around 20 minutes, with short 1-2 minute clips for quick topic views. Today: episode #128.

Subscribe to either or both on michaelparekh.substack.com. They run as separate Sections you can opt into or out of.

(NOTE: The discussions here are for information purposes only, and not meant as investment advice at any time. Thanks for joining us here.)