Mark Zuckerberg’s daughter told him she wanted to be like Taylor Swift. His answer: “You can’t. That’s not available to you.” He had taken August to a Swift concert. Afterwards, she told him she wanted to be like the singer when she grew up. Most parents would probably smile and encourage the dream. Zuckerberg gave her that answer instead. In the interview below, he tells the story himself. And his daughter’s response is the part worth hearing. August thought about it, then decided she wanted to grow up into someone other people would want to be like. Someone called August Chan Zuckerberg. Her father’s reaction: “Hell yeah.” Coming from Zuckerberg, that exchange is easy to admire and just as easy to question. There is ambition in wanting to become someone others look up to. There is also a lot of pressure in turning a child’s excitement after a concert into a conversation about becoming exceptional. She went to see her favorite singer. She came home talking about making her own name. Would you have answered your daughter the same way?

Why would a cloud company hand a customer up to 5 percent of itself? Akamai has been moving a big share of the internet's traffic since 1998, and most people have never heard of it. Last week it announced that Anthropic will spend $11.6 billion on Akamai's cloud over seven years. In the same announcement it gave Anthropic a warrant, the right to buy up to about 5 percent of Akamai later at a fixed price. We want to slow this one down, because the headline number and the money Akamai actually collects are two different things. Akamai expects nothing from the contract this year and reaches around $1.7 billion a year only from 2028, at full speed. For a company that brought in about $4.2 billion in 2025, that is a long wait. Akamai spends first. About $5.5 billion goes into new capacity through 2028, roughly 47 cents of every contract dollar, mostly out the door before the big payments arrive. Now the 5 percent. A warrant pays off only if the stock rises above its set price, and Akamai set that price at $111.33, about where the stock traded before the news. For comparison, when AMD gave OpenAI a similar right in October 2025, the price was one cent a share, which is close to handing over the stock outright. Most of the 5 percent also depends on Anthropic committing more money. 40 percent unlocks at the first payment, and each extra $3 billion unlocks another slice. The stock jumped 22 percent after hours on the news. Six days later it closed at $106.35, under the warrant's own price. We read the 5 percent as a bet Akamai is making on itself, paid for with Anthropic's future spending. So far the stock sits below the warrant price, and exercising the 5 percent today would pay nothing. We think most people are watching the wrong hardware. What Anthropic is buying is ordinary CPUs on Akamai's distributed network, the unglamorous capacity AI agents burn through, rather than the GPU clusters that get the headlines. We expect people who can price, run and sell that kind of capacity to stay scarce for years. That window has a date on it. Akamai's own plan has this contract at full speed in 2028, and by then the CPU side of AI is either a real market or a $5.5 billion build-out with one big tenant.

What happens to a habit when the study behind it gets retracted? For years I've cut every big project into small deadlines I set for myself. They live on a sticky note at the bottom of my monitor, one line per Friday, like a chore chart on a fridge. Part of the reason was a 2002 paper by Dan Ariely and Klaus Wertenbroch. Self-imposed deadlines help, evenly spaced ones help more. I had assumed that part of my week was settled science. This September the journal retracted it, at Wertenbroch's own request. The data sleuths at Data Colada found that in one experiment, 18 of the 20 people in one group had a duplicate elsewhere in the data, down to identical proofreading scores. Ariely's own advice about the finding is now to believe it less. Turns out the habit didn't break. That's the annoying part. I looked at the sticky note after reading the notice and the Friday lines were still there, still working on me. The study was never holding the habit up. It was the receipt I kept in the drawer after I'd already bought the thing. What still stands is smaller. People really do choose to bind themselves early, and that part has held up. Whether evenly spaced deadlines make the work better is another matter, and a new replication in the same journal found no evidence that they do. Ulysses figured out the binding part long before any lab did. He had his crew tie him to the mast so he could hear the sirens without being able to sail toward them, and nobody ran a controlled trial on the rope. The sticky note is still on the monitor. This Friday's line says finish the draft, and I haven't decided yet whether I believe it.

I'd keep the deadlines and check whether they improve the work or just make Friday feel urgent

The hedge fund manager who taught Michael Burry to pick stocks accidentally destroyed the mutual fund industry in a free Talks at Google lecture on the two-line formula that turned $10,000 into $8.3 million. His fund charged 2 and 20. He gave the strategy away in a $12 book. Almost no one paying a financial advisor 1 percent of their retirement has finished the lecture. His name is Joel Greenblatt. He founded Gotham Capital in 1985 and compounded roughly 40 percent a year for twenty consecutive years. He returned all outside capital in 1994 because his edge was too crowded to scale. He has been teaching value investing at Columbia Business School every year since 1996. Michael Burry - the doctor who shorted the housing bubble in "The Big Short" - learned to pick stocks from Greenblatt's 1997 book. The 55-minute clip in this video is Greenblatt at Google in 2017 walking a room of engineers through his Magic Formula. The whole framework fits on one napkin. Rank every stock in the S&P 500 by return on invested capital, highest first. Then rank every stock by earnings yield, highest first. Sum the two rankings. Buy the top twenty. Hold for one year. Sell. Repeat. That single formula would have beaten the S&P 500 by roughly 14 percent a year over the last three decades if any retail investor had actually followed it. "Cheap and good beats expensive and average. Every time." That is Joel Greenblatt at Google in 2017. He has repeated the sentence in every public talk since. Almost no retail investor buying Nvidia at $140 has heard it. Every hedge fund on Wall Street pays $500,000-a-year analysts to backtest more sophisticated versions of the same equation. Every financial advisor in America charges you 1 percent of your account per year to underperform it. The Talks at Google video is free on YouTube. "The Little Book That Beats the Market" is twelve dollars on Amazon. Almost none of the millions who watched have ever run the two-line formula on their own portfolio. The framework is free. The willingness to actually run it before your next stock pick, retirement rebalance, or brokerage transfer is the entire edge.

Can a habit go backwards while you're still doing it? I had assumed habit formation ran one direction. You repeat something, it gets easier, eventually you stop deciding. That was the model I'd been running on for years and I never once checked it. Turns out some habits run the other way. A reanalysis published in Personality and Social Psychology Bulletin in April 2026 went back over six months of data from 254 people. Every month those people had rated how automatic their own behavior felt, and the team tested ten things that might explain why automaticity rises or falls. Burgler, Milyavskaya, McMillan, Burns, Lally and Gardner, if you want to go find it. They then grouped people by how their automaticity moved over the six months. Two paths came out instead of one. Some people's behavior was getting more automatic. Some people's was getting less automatic, while they kept doing the thing. Six of the ten explanations came out significant. Wanting to do it rather than having to. Feeling energized. Feeling pleased. How hard it felt, which pushed automaticity down instead of up. How often you did it. And whether it happened in the same place at the same time. Four of those six are about how it feels while you're in it. One is frequency. One is context, and context is the one I'd been ignoring for a decade. I used the word discipline to explain why one of my habits kept sliding while the others held. Going by this data the real difference was probably that the sliding one had no fixed place in my week. Which is the annoying part, because that's a calendar problem and I spent years treating it as a moral one. Discipline isn't on the list of six at all. I think that's because discipline was never a driver. It's the bill you pay when the drivers have gone missing. The gap is obvious once you see it. Every habit app counts streaks, which is frequency, the one driver everybody already measures. Nothing tracks whether your context is still holding. That's a product. What I can't tell you is how many people ended up on the downward path. The paper's behind a paywall and the abstract doesn't give the split. So I'm reading a shape, not a size. I moved that habit to a fixed slot in August. Too early to say. But I've stopped calling it character.

You stall on easy work because nobody will keep the result. I called it discipline for about ten years. It's not discipline. It's my head pricing the job before I start it, and the price goes up when the output is going to disappear. The pattern is narrow and I can name it. I stall on tasks I'm completely capable of doing, and they're never the hard ones. A summary nobody will open. The draft written to be replaced. Calls that end with no note. Yilin Liu and Fan Yang published something in July in Psychological Science that describes the mechanism better than I can. Four experiments, 582 adults and 604 children between four and nine. The adults got two options. Build a figure, which takes effort, or tip the pieces into a bin, which takes none. When the finished build was put in a box and kept, 83% chose the effortful one. When it was taken apart in front of them, 34% did. Same task, same room. The only thing that changed was what happened to the result afterward. What got me is the size of the threshold. It isn't meaning or impact. It's a box. The thing gets set aside intact instead of pulled apart, and that's enough. Liu said what surprised her most was how little significance turned out to be necessary. They tried the obvious fix too. One condition emphasized that the task would be fun, and it only partly closed the gap. Making work more enjoyable doesn't repair work whose output gets erased. And the effect shows up at four years old, which kills my favorite explanation. I'd assumed this was something offices did to me in my thirties. Yes, this sits next to the IKEA effect, where you overvalue what you assembled yourself. It's not the same thing. In a second experiment, people who built the figure gave up a sticker to keep it, and people who only watched someone else build took the free option and let it be dismantled. The value isn't in owning it. It's in having made it, and then needing it to still exist. The authors name their own limit in the paper, and that limit is the entire modern problem. Every task used physical objects. Nobody tested the digital version. Almost everything I produce is a file, and a file is dismantled by default. No box. No shelf. Nothing that says this stayed. What I've started doing is small and probably not enough. One result a week gets a permanent form, printed or saved somewhere I never clean out. A box, basically. It also changed how I read a job offer. What it pays is one question. What survives it is a separate question, and I suspect the second one bills you later, in the things you can no longer start. One thing I can't settle. Does a public file count as a box, or will this only work with something my hands can pick up?

Maybe the box isn’t physical. It’s just knowing your work won’t vanish the second you’re done with it.

This landed. I kept calling it discipline too, then realised I only stall on the stuff that gets overwritten the same day — notes no one opens, drafts that get replaced, calls with nothing written down after. The box vs bin thing is almost too simple, but it explains why “make it more fun” never worked for those tasks. Started printing one thing a week and putting it in a drawer I don’t empty. Doesn’t fix the whole problem, but the resistance dropped. Still not sure if a public file counts as a box though.