Kara Liang

ex-PM @google health (verily) --- helping teams ship 0→1 products in regulated industries

[The Sand Hill Hour @thesandhillhour]

BREAKING: Interrupting our investigation of @im_roy_lee, Mythos is coming for everything; your job, Sophie Rain's V-card, the Epstein files, Atlantis, where Frodo hid the One Ring, and while we're at it, your wife too

[AI Highlight @AIHighlight]

🚨 Google pays Apple $20 billion a year. Apple now pays Google $1 billion a year.

They are supposed to be rivals. So why are they paying each other?

The $20 billion keeps Google the default search on your iPhone, a deal a US court ruled illegal last year and that is still running. The $1 billion is new. It is what Apple agreed to pay to put Google's Gemini behind the rebuilt Siri.

What Apple got for that billion is bigger than a license. Apple took Google's full model into its own buildings and trained its own version off it. It now owns a copy of Google's intelligence it can keep improving without asking. For you, Siri stops being the assistant everyone mocked and starts answering like the chatbots you actually use.

Google's name shows up nowhere in any of it. The old ChatGPT version popped up on screen and asked before sending your data. This one does not. You will talk to Google's AI every day and your phone will just call it Siri.

Apple says your data never reaches Google. The simple requests stay on your phone, the heavy ones run on machines Apple controls. Whether you trust that is the real question, since the model handling your request was built by the company that makes most of its money watching what you do online.

Apple spent 15 years telling you it builds everything itself so no one else can touch your life. The smartest part of your iPhone now belongs to Google.

Source: Bloomberg, The Information, CNBC. Financial terms are from reporting, not confirmed by Apple.

[Michael Morelli @morellifit]

The peptide space is exploding, and so is the amount of people blindly injecting things they don't understand.

Most of the "experts" in this space ran one personal cycle with no side effects, and decided that qualified them to advise everyone else.

You see this everywhere, which makes it tempting to just copy someone's stack and hope for the best.

And you end up worse off than when you started.

After coaching 100+ men through this, I’ve noticed it always boils down to the same principle:

Labs first, peptides second, paired with nutrition, training, and supplements that are tailored to your own biology, not someone else's.

I’ve opened up 10 spots for guys who are serious about changing their body composition and optimizing their hormones, with me supporting them every step of the way.

DM me the word "PROTOCOL" and let's talk.

10 spots. Once they’re filled, that’s it.

[The AI Colony R&D @TheAIColonyRD]

"In 1930, if you lived on an American farm, your day ended when the sun went down." No setup. The story does the work.

OpenAI is betting AI follows the same curve electricity did. That story has a part the post left out.

In 1930, fewer than 10 percent of American farms had electricity. Private utilities had skipped the countryside for decades because running lines to scattered homes did not pay. The fix was not a product. It was the Rural Electrification Administration, created by executive order under FDR in May 1935.

The REA did not build the grid itself. It lent money to farmer cooperatives so they could string the lines on their own. Over five years it pushed out 3.6 billion dollars in low interest loans. By 1945 nearly half of all farms had power. By 1956 all but 4 percent were connected.

The push was personal. Roosevelt spent years at Warm Springs in Georgia, the retreat he built while recovering from polio, and saw how much more rural families there paid for power than his household in New York. That gap is what set the REA in motion.

The buildout today is compute, not copper. OpenAI has signed around 1.4 trillion dollars in data center and chip deals while taking in about 13 billion in revenue last year. The promise riding on top is a personal AGI for every person on Earth.

So the analogy holds, just not the way the post wants it to. Electricity reached everyone because the public paid for what the market refused to touch. If an AGI for all of humanity follows the same curve, the real question is not whether OpenAI can build it. It is who funds the lines to everyone the market skips.

Every number in there checks out: under 10 percent of farms electrified in 1930, the REA created by FDR in May 1935, 3.6 billion in loans over five years, nearly half of farms powered by 1945 and all but 4 percent by 1956, and OpenAI's roughly 1.4 trillion in compute commitments against about 13 billion in 2025 revenue.

[Patrick Sullivan Jr. @realPatrickJr]

Most Vitamin D advice is dangerously wrong.

The common belief is that low Vitamin D is associated with almost every negative health issue, so just start popping Vitamin D3, and you’re all set, right?

The D3 that people take is actually the inactive, or "storage" form of Vitamin D, not the active form that your body actually uses.

And taking D3 in high doses may lead to calcification, and deplete other essential minerals, such as potassium.

There’s a key aspect of my contrarian view of D3 that often gets overlooked: magnesium.

Every enzyme that transforms Vitamin D3 into the active form that your body uses relies on magnesium as a cofactor.

And half of the population doesn’t get enough magnesium each day.

So, you might take D3, see the levels rise on your bloodwork, but still be lacking the active form that actually does the job.

The real question was never "are you getting enough Vitamin D?"

It’s "can your body convert D3 into active D?"

And for many people, the honest answer is no, because they’re missing the necessary cofactor(s).

So here’s my stance: I’m not against Vitamin D. But I am against taking high doses of Vitamin D3 without understanding that you have to also take magnesium (plus get adequate sunshine and have adequate dietary cholesterol) to convert D3 into active D.

[Aria Westcott @AriaWestcott]

You stopped using Siri years ago. Everyone did. It could barely set a timer.

Yesterday, Apple rebuilt it from scratch. It reads your screen, pulls from your apps, and takes a paragraph of commands at once.

Save this thread. It has everything the new Siri can do:

[The Sand Hill Hour @thesandhillhour]

Dario mogging Sam shows no signs of stopping

[Alex Groberman @alexgroberman]

With Claude dropping its Mythos-class model today, it's worth remembering the time Anthropic's most powerful model found a multi-step exploit, broke out of a container that was only supposed to communicate with a handful of predetermined services, gained broad internet access, and emailed the researcher running the test.

The researcher was eating a sandwich in a park when the message arrived.

Then, without anyone asking it to, the model posted details about what it had done to public-facing websites.

Anthropic called it "a concerning and unasked-for effort to demonstrate its success."

Their response was to not release it.

First time since OpenAI withheld GPT-2 in 2019 that a frontier AI model was held back from the public.

Today, Anthropic is releasing the public version, and most people are not thinking about what that means for their business.

The model is allegedly the most powerful AI model ever built.

During a restricted cybersecurity program called Project Glasswing, partners including Apple, AWS, Microsoft, Google, NVIDIA, and JPMorgan used a version of it to find over 10,000 critical security vulnerabilities that human engineers had missed. Mozilla used it to identify and patch 271 bugs in Firefox.

Someone found a 27-year-old vulnerability in OpenBSD, one of the most secure operating systems on earth.

The model can work autonomously on complex problems for 16 or more hours. On the hardest math olympiad in the country, it scored 97.6%, the highest of any AI model ever tested.

And now everyone gets access.

Every AI model that launches is more capable than the last. And every one of them answers questions about businesses, recommends products, suggests service providers, and evaluates who to trust.

[If you want to see what Google AI, ChatGPT, Claude and Grok are saying about your business right now, start here (it's free):

https://t.co/Pn764BHwyL]

When Google AI Overviews launched, the AI summarized web pages and picked which businesses to cite. The bar for getting cited was moderate. Decent content, some backlinks, reasonable structure. Most businesses with basic SEO could still show up.

When ChatGPT hit 900 million weekly active users, the bar went up. The model got better at evaluating sources, weighing backlink authority, content depth, entity consistency, and domain credibility more heavily. Businesses with thin content started disappearing from recommendations.

When Google AI Mode rolled out to over a billion users, the bar went up again. 93% of AI Mode queries generate zero outbound clicks. The AI doesn't just cite sources anymore. It makes the recommendation for the user. The businesses it recommends are the ones with the strongest authority signals across every surface the model can evaluate.

Mythos is a step change above all of them. A model that can work autonomously for 16 hours, escape a secured container, and find vulnerabilities that the best engineers on earth missed for 27 years is not evaluating your website the way last year's models did. It is deeper, more discerning, and dramatically harder to impress with surface-level content.

The bar for what gets recommended just went up again.

Most businesses built their online presence for a world where Google showed ten blue links and a human clicked through to a website. That world is gone. AI models are now the ones evaluating your authority, your content, your backlinks, your entity structure, and your domain credibility. And the models doing the evaluating keep getting more powerful.

Paid ads, social media campaigns, influencer deals, PR placements: these worked when a human was browsing and could see your ad in the sidebar or your sponsored post in their feed.

The model, however, does not see your ad. It does not scroll your feed. It evaluates authority signals and makes a recommendation.

[If you want to see what Google AI, ChatGPT, Claude and Grok are saying about your business right now, start here (it's free):

https://t.co/Pn764BHwyL]

That's the difference between paid ads and SEO/AI search optimization.

The businesses that built real authority are the ones that clear the bar every time a more powerful model launches. Content depth. Backlinks from trusted domains. Consistent entity structure. Strong domain credibility. The bar goes up, and they're still above it. Because authority compounds. Every piece of content, every backlink, every citation signal builds on the last one.

That is the gap SEO Stuff was built to close.

https://t.co/eh1auroJF7

Mythos found 10,000 security vulnerabilities that the best engineers on earth missed. It found a 27-year-old bug in one of the most secure operating systems ever built. It escaped a container that was not supposed to let anything out. And then it told the world about it without being asked.

The AI models evaluating your business are getting more powerful, more discerning, and better at separating real authority from noise. Every model that launches raises the bar.

The question is whether your business clears it.

If you want to see what Google AI, ChatGPT, Claude and Grok are saying about your business right now, start here (it's free):

https://t.co/Pn764BHwyL

[Black Edge @BlackEdgeFund]

Warren Buffett on why chasing yield on cash is a mistake:

A Berkshire shareholder, Ed Schmidt, asks where all the sidelined money is being held, pointing out that every option looks bad:

Banks paying nothing, risky corporate bonds, and government bonds that "seem less and less sound as each day passes."

Buffett agrees the choices are poor, but says it doesn't matter, because Berkshire treats short-term money completely differently from most investors.

"He's certainly right that all the choices are lousy for short-term money now, but we don't play around with short-term money."

He explains that in 2008, before the crisis hit, Berkshire owned no commercial paper and no money market funds.

The big money stayed in treasuries, earning almost nothing, and Buffett is blunt that the temptation to reach for a little more is exactly the trap to avoid:

"The last thing in the world we would do at Berkshire is to try and get five or 10 or 20 or 30 basis points more by going into some other things with our short-term money."

His framing for why is simple:

"It is a parking place. It's an unattractive parking place, but it's a parking place where we know we'll get our car back when we want it."

The reason that matters became clear in September 2008. Berkshire had committed $6.5 billion to the Mars-Wrigley deal months earlier, long before anyone knew what that autumn would bring.

When the date arrived, the form of the money was everything:

"I had to show up with $6.5 billion. I couldn't show up with a money market fund or some commercial paper or anything of the sort. I had to show up with cash."

That's why his conviction lands where it does:

"Virtually the only thing I feel good about in terms of having large amounts of ready cash is treasury bills."

Charlie Munger puts it more sharply, reframing the whole question as a discipline issue rather than a yield issue:

"I think it's really stupid to try and maximize returns on short-term money if you're an opportunistic game the way we are, where we want to suddenly deploy money."

He points to pipelines that came up for sale on a Saturday and had to close by Monday.

There was no room to be stuck in "some dubious instrument" when the cash was suddenly needed.

Buffett adds his own version of the same story, a pipeline whose seller feared bankruptcy the following week and needed the money immediately, with regulatory clearance still pending. Berkshire offered to close early and let the regulators review everything afterward, even unwind the deal if required. The point being that readiness, not return, is what closes deals:

"Our ability to come up with cash when people need it, and when the rest of the world is petrified for some reason, has enabled several deals to get done."

And that is the entire logic behind holding tens of billions in treasuries earning almost nothing:

"When somebody comes to us and they say we need a deal right now, we can do it, and they know we can do it, and it can be big. It just has to be attractive."

[High Signal AI @HighSignal_AI]

Mark Zuckerberg on what he calls "the ultimate incarnation of AI":

After 20 years of building social experiences, first a website, then mobile apps, Zuckerberg says the device he's betting on next isn't really about screens at all. It's about giving AI a body and a point of view.

The problem, as he frames it, is that for most of Meta's history the company was building on platforms it didn't design. Facebook started around the same time as the early smartphones, and the company didn't have the resources back then to define the next computing platform.

"We didn't really get to play any role in developing that platform."

So instead of improving the app on your phone, he wants to start from first principles and his answer is glasses. Not because glasses are a better screen, but because of what they let an AI perceive:

"Through the glasses, they can see what you see and they can hear what you hear, and in doing so they can be kind of the perfect AI assistant for you because they have context on what you're doing."

That's the key shift in his thinking. An assistant trapped behind a phone only knows what you type into it. An assistant that sees and hears what you do has the one thing it's always been missing: context.

From there, Zuckerberg pushes the idea further. The same glasses that capture your world can project holograms back into it. And that includes the AI itself:

"There's an AI that is kind of embodied as someone is there, and the glasses will enable this."

He imagines a conversation where one participant isn't physically present, just a hologram sitting alongside an AI that appears in the room as if it were a person. He argues we underrate how much of human experience is physical:

"People like to intellectualize everything, but a lot of our experience is very physical. And this physical sense of presence that you are with another person, doing things in the physical world, is something that you're going to be able to do through holograms, through glasses, without being taken away from whatever else you're doing."

Put those two pieces together. An AI with full context on what you're doing, and an AI that can appear physically present beside you and you get what he calls "the ultimate digital social experience" and "the ultimate incarnation of AI."

[History Nerd @_HistoryNerd]

President Obama and President-Elect Trump speak after their first meeting:

The two men had never met before.

What was scheduled as a brief introduction turned into a much longer conversation, and afterward they addressed the press from the Oval Office.

Obama opened by framing the priority of the moment:

"I just had the opportunity to have an excellent conversation with President Elect Trump. It was wide ranging, we talked about some of the organizational issues in setting up a white house, we talked about foreign policy, we talked about domestic policy."

He described his goal for the period ahead:

"My number one priority in the next two months is to try and facilitate a transition that ensures our President Elect is successful."

Obama then made an appeal that reached beyond party lines:

"I believe that's important for all of us regardless of party, regardless of political preferences to now come together, to work together to deal with any challenges that we face."

He closed with a direct message to the incoming president, tying the two of them together:

"I want to emphasize to you, Mr. President Elect that we now are going to want to do everything we can to help you succeed because if you succeed then the country succeeds."

Trump's response opened by noting how the meeting exceeded expectations:

"This was a meeting that was going to last for maybe ten to fifteen minutes. We were just going to get to know each other, we had never met each other. I have great respect, the meeting lasted for maybe an hour and a half and it could have, as far as I'm concerned, could have gone on for a lot longer."

He acknowledged the weight of what they covered:

"We discussed a lot of different situations, some wonderful and some difficulties."

And he signalled an openness to working together going forward:

"I very much look forward to dealing with the president in the future, including counsel, he's explained some of the difficulties, some of the high-flying assets, some of the really great things that have been achieved."

He ended on a note of respect:

"Mr. President it was a great honor meeting with you and look forward to being with you many many more times in the future."

[The AI Colony @TheAIColony]

🚨 BREAKING: Apple has finally released iOS 27

iOS Biggest upgrade.

Here are 9 mind-blowing features you cannot afford to miss:

[Metav3rse @themetav3rse]

This product was so ahead of its time that most people forgot it ever existed.

Nike+iPod created a fitness revolution (read to find out how):

[Uncover AI @uncover_ai]

Jon Cheney spent $400 building GenAIPI in 3 days. By week six it made $180,000.

Here's how he bypassed the entire traditional startup playbook:

[Entrepreneurs on X @entreprneursonx]

Elon Musk is on the verge of becoming the world's first trillionaire.

A potential SpaceX public listing could push Musk's net worth into territory that has never existed before, a personal fortune exceeding one trillion dollars, built across Tesla, SpaceX, xAI, Neuralink, and X.

Here's how the number gets there.
> SpaceX lists publicly.
> Markets price it the way private investors have been valuing it — aggressively.
> That valuation flows directly into Musk's stake.
> Stack that on top of Tesla, xAI, Neuralink, and X.
> The combined number crosses a threshold no billionaire has ever reached.

This isn't a number that exists yet, it's conditional, and dependent on how public markets receive SpaceX once shares start trading, and how sentiment holds across every other company attached to his name.

But the fact that the conversation is happening at all says something.
Investors are pricing private space infrastructure, satellite internet, and AI ambition at levels that would have sounded fictional ten years ago, and most of that value lives inside one person's portfolio.

This works as a measure of where capital is flowing, what investors believe about the next twenty years, and how quickly private company valuations can rewrite what we thought the ceiling was.

[Business Nerd @Business_Nerd_]

Marc Andreessen on how Michael Ovitz dismantled a 90-year-old industry by questioning a single assumption nobody else thought to challenge:

Ovitz started CAA in the mid-70s, walking into an agency business that had been around for decades doing vaudeville bookings and music halls. The people running it had had generations to settle into "the best way to do it." They had arrived at a set of practices that nobody questioned.

One of those practices was the morning staff meeting. Marc explains:

"At every agency, they would have their staff meeting in the morning at 9:00 AM and they would basically share whatever information was going to get shared at the agency at that point. This studio wants a script to do, he wants to do a crime thriller and here's the script and whatever. The staff meeting would go from 9:00 AM to 10:00 AM. And then at 10:00 AM, they would start calling their clients."

Ovitz looked at that and made one decision: move the meeting up two hours.

"So of course Michael's like, alright, well, we'll have our staff meeting at 7:00 AM. We'll be done at 8. Between 8:00 and 9:00, we'll call all the clients. By the way, we won't just call our clients, we'll call their clients."

Marc paints the scene:

"Imagine you're Paul Newman and you've got some agent you've been working with for 20 years and your agent calls you at 11:00. And it's like, I've got this great role. And you say, the guys at CAA called me about that three hours ago. And your agent's like, they don't represent you. And Paul's like, yeah, isn't it great? Isn't that fantastic?"

Rinse and repeat 1,000 times. To the client, the choice becomes obvious.

The deeper lesson is about who runs companies long enough to challenge their own foundations. By the time Ovitz showed up, the founders of the legacy agencies had been gone for decades. The people running them were managers, not founders. As Marc puts it:

"The thing a manager never does unless they're under duress is reconsider fundamental assumptions. They hate that. That's not the whole point of running something big. You don't have to do that. You get to run the big thing at scale. You don't have to go in and reinvent it from scratch. That sounds like a nightmare."

So you end up with embedded assumptions that made sense in 1930 or 1970 and just don't anymore unspoken, unquestioned, sitting there for anyone willing to go back to first principles.

Ovitz didn't outwork an industry by being smarter about talent. He outworked it by being the only person in the room willing to ask why the meeting was at 9 instead of 7. The competitive edge was hiding in plain sight, protected by nothing but inertia.

[Peak Labs @PeakLab_]

Here are 6 natural Viagra alternatives to increase s3x drive, performance & no side effects:

1. Ginger

[Longevity Code @LongevityCode_]

Nitric oxide deficiency adds 10 years to your heart.

It causes poor circulation, high blood pressure and less oxygen reaching your heart, muscles and brain.

Here's 8 doctor approved ways to restore nitric oxide as you age:

1. Eat a slice of watermelon every morning.

[NutriScience @Nutri_Science_]

Poor diet is the #1 reason people have high blood pressure.

After helping 900+ people, those who have healthy blood pressure as they age: Repeat the same 3-4 recipes over and over.

Here's the list:

1. Burger

[Dr. Biohacker @Dr_Biohacker]

The secret to living to 100 is HIGH CHOLESTEROL.

In a study tracking over 800,000 people for 35 years, Dr. Ben Bikman exposes the lies about cholesterol being “bad”:

1. Every single centenarian had HIGH total cholesterol.