SuccessDecoded

In 2010 Jobs showed a laptop touchscreen onstage and called it terrible. Apple holds its Pro and Tell event in Los Angeles today, and the expected star is its first touchscreen MacBook Pro, according to Bloomberg and Macworld. Jobs is about to get quoted a lot. At Back to the Mac in October 2010, he said Apple had done tons of user testing and a vertical touchscreen wears you out until your arm wants to fall off. That's the part people have repeated for 16 years. Seconds later he gives the rule: "Touch surfaces want to be horizontal." Then he goes straight to the multi-touch trackpad, which is where Apple actually put touch on the Mac. My reading, which he never spelled out: Jobs's objection was tired arms. A laptop screen is still vertical, so his own test still applies, and Apple has to pass it. A manager who writes the reason into a no, the way Jobs did, can reverse it years later by showing the reason no longer holds. So the question for every review tonight is the one Jobs wrote: after a while, does your arm want to fall off?

the test now goes to the first review that mentions arms

Who fixes the culture after AI deletes the meetings? Eric Schmidt sat down with Nicolai Tangen, who runs Norway's sovereign wealth fund, and gave Fortune its story. He used to sit in meetings 12 hours a day, and AI, he said, can strip out that coordination tax. A 10,000-person company probably has a couple of thousand people it doesn't need. Asked how AI reshapes business, he said almost none of the companies listening "will actually do the necessary changes in their culture." Companies are organized around belief systems, he explained, where the morning is predictable and everyone is comfortable. You'd sit through 13 minutes of regulation talk to reach it. The meetings quote Fortune ran comes 23 minutes later. This is my reading; he never said it. The cut is the easy part. A friend of his laid off 20 percent of his people, mostly the glue people who did planning and shuffled numbers, and got death threats for it. A company can do exactly that and still run the old approval chain with fewer hands. So the scarce skill is a manager willing to make the morning unpredictable. A team lead who deletes the weekly status meeting but keeps the same three sign-offs gets the layoff without the speed. Plenty of the people cheering the coordination tax line work at exactly the kind of company he says won't change.

Jon Stewart used a nesting doll to take apart the Russian diesel deal. Diesel was running about $6.28 a gallon in early October, up from $3.68 a year before. Then the US struck a deal for Russia to ship more of it to American and world markets, three weeks after signing a law to punish anyone who buys Russian oil. What everyone repeated from his Daily Show segment was the punchline about Putin's table. What he actually did on stage was quieter. He took that one decision and turned every justification for it into a separate wooden layer, then opened them in order: first the promise of lower prices, then the claim that this pressures Russia. Russia ships 300,000 tons right away and 500,000 in November, with more promised after that, announced October 9. The sanctions law was signed September 18. On October 16 a single executive order switched on the law's new tariffs and switched them off again. Each doll is smaller than the last. At 1:08 he says it plainly, "a physical demonstration of the many layers of stupidity." Moments later a reporter hands him the next layer: 300,000 tons would have "very little effect" on prices long term. He never names the trick, so this part is my reading. When a decision is incoherent, he doesn't argue with it. He makes each excuse a physical thing and lets the audience watch it shrink. A buyer who lays out each reason for a purchase on its own card, biggest first, often finds the deal resting on the last and smallest one. A seller can run the same check on a pitch before a customer does it for them. Diesel is what moves groceries and farm equipment, so the families paying $6 at the pump end up holding the smallest doll. Which decision at your own job would fall apart if someone opened it one layer at a time?

Li Lu, the one outside manager Munger trusted, warns against chasing fund size. Li Lu, who runs Himalaya Capital, spoke at Columbia this month. The write-ups that got passed around were about AI needing a human in the loop and a stock market with a casino attached. The useful part sits in the Q&A, when a man two years into value investing asks about the risk of lagging the market while he waits. You'd have to sit through more than an hour of talk, much of it on China and AI, to reach it, about 1:16 into the Columbia Business School video. He walks down the reporting ladder first: good numbers every year, then every quarter, every month, every day, and pretty soon you become a speculator. Then the line: "you cannot really say I want to have the biggest fund." This is my reading; he never said it this way. A fund grows out of the clients who stay through a bad year, and you can't set a target for loyalty the way you set one for size. Flip it and it says how often you report decides what you buy. Promise a number every quarter and you'll start buying whatever looks good by quarter's end. Outside a fund it looks like a sales manager who stops sending weekly numbers up the chain and reports twice a year. She will look slow for a while. Li Lu had Charlie Munger's $88 million in his fund from 2004 to look slow with. Many people in that room answer to clients who check every month, so the advice fits best the manager who no longer has to pick up that call.

Six minutes with Jim Cramer, and Elizabeth Holmes never once mentioned her machine. If you've seen her on Mad Money, you remember him comparing her to Steve Jobs. That moment still gets replayed as an awkward rewind. She puts a Nanotainer on the table, a tube small enough to pinch between two fingers, built to swap the vials from your arm for a drop from your finger. The analyzer that was supposed to run the tests never comes up. Then Cramer asks the only hard question of the segment: two lab giants with billions in revenue, so why can't they crush you? Her answer: "this is not about us or our technology." Then she moves on to health access as a basic right and to Medicare prices. My reading of the tape: the one question about whether the business could survive got an answer about why it matters, and why it matters can't be fact-checked on live TV. A tube can be photographed. A test result has to be right. On that same set she said lab results drive 80 percent of clinical decisions. That October the Wall Street Journal reported, citing a former senior employee, that only 15 of the 240-plus tests Theranos offered ran on its own device. The count was from the end of 2014, months before the April 2015 interview. "You Can See Everything," the new documentary about her, opened to about $15 million last weekend, and her old interviews are circulating again. A hiring manager can borrow this for free. She writes down which of her questions got a mission back instead of a mechanism, and that's the one she asks a second time. The mission was the only part of Theranos that never had to work.

New hire from big tech walks in unbriefed and rebuilds their last employer. The line that traveled from Bloomberg's Anthropic series in June was the org chart. Dario Amodei has one direct report, and the business press ran with it. In the extended cut, Emily Chang asks how a company keeps its values at this scale. He says the leaders' values aren't the risk. The risk is that the people inside change very fast. Hire from big tech without telling people how Anthropic operates, he says, and "they'll simply recapitulate the only thing they know." That's why he spends probably half of his time talking to the company about culture. This is my reading; he never said it. Values rarely leak out of a company. Habits leak in, because every new hire arrives carrying a working manual from the last job, and silence means that manual wins. Read that way, the one direct report stops looking like a management quirk. It's what spending half of his time on culture costs a CEO. A manager who hires three people this quarter and writes down how her team actually makes decisions has done the cheap version of that half, without his budget. Writing that manual eats the hours a founder wanted for the product. It never feels urgent until the third hire runs your Monday meeting the way their last boss did. Who wrote the manual your newest hire is running right now?

Why would AI labs say no to $200 billion in revenue? Everyone clipped the headline from Dylan Patel's August sit-down with Dwarkesh Patel: two labs will soon control most of the world's compute. The $200 billion was Dwarkesh's own math, for labs that keep adding training compute instead of selling it. Dylan Patel's answer is the part worth a third listen. He says the labs will put less and less of their compute toward serving customers over time, and he flags it himself as very non-consensus. Selling tokens, in his words, exists "so you can grow your training fleet." You'd have to sit through half an hour of fab capex and chip prices to reach it, in the chapter that starts at 29:43. At $30 to $40 million of revenue per megawatt, a lab today gives about 40 percent of its compute to inference, by his numbers. At $60 to $70 million, he asks, do you pay dividends, or do you go build AGI? This is my reading. He never said it in one line. For these labs revenue is the fuel bill, and the customer is financing the next model without being told. Copying it is expensive for anyone smaller than OpenAI or Anthropic. A startup that plows every sale back into its next model looks less profitable than its rivals for years. It only pays if the next model is worth more than every dividend it skipped. Flip it, and on his logic the faster these labs grow, the smaller the share of their compute that works for the companies paying them.

Why did Jamie Dimon go through every Bank One loan himself? The bet is the part people repeat. Right before taking the job he bought about $60 million of Bank One stock, roughly half his money. The part he rushes past comes later. Bank One was booking money-losing loans as profitable, he says, so "I went through every single loan in the books." You'd have to sit through about 14 minutes of career stories to reach it, in the chapter that starts at 4:50. Once he saw how the loans were booked, he marked every one down and told the board. He had just put half his money into that stock, and his own markdown made the bet look worse. That's my reading; he never said it. The numbers back the story. His letter to shareholders says Bank One lost $511 million in 2000, after about $5 billion in charge-offs, reserve additions and write-downs. Most people file this under risk management. Flip it and it says something less comfortable about anyone who takes over a team. The price is real for anyone without his cushion. A manager who rechecks every account she inherits spends her first quarter reporting worse results than her predecessor did, and she has to survive that quarter. He refused to inherit a number he hadn't checked, even when the unchecked number was flattering him.

SpaceX's president stands on a stage and calls her workhorse rocket a minivan. Elon Musk joined that All-In Summit session in September partway through, and the write-ups followed him. The line that mattered came from the person who runs the company day to day. Gwynne Shotwell, who had the stage before him, spent part of that session on her own product line. "A Falcon 9 with a Dragon capsule on top is like a minivan," she said, and the point was Mars. Six months is a long road trip, and that ride is Starship. Taken alone, the line sounds like SpaceX's president retiring her workhorse in public. That's not what happened next. Less than three weeks later, a Falcon 9 and Dragon carried NASA's Crew-13 to the space station. The minivan didn't get parked. It got demoted in a sentence while it kept flying the people who pay. This is my reading; she never said it. You can call your workhorse obsolete out loud only while it's still flying paying customers, and only because something else pays for it. SpaceX went public in June, and its April to June numbers show who pays. The rocket segment lost $542 million from operations that quarter, while Starlink earned close to $1.7 billion in operating profit. A founder with one product and no Starlink can't afford that line yet. A software founder would announce the new version only once the old one's renewals were covered elsewhere. Until then the founder would keep the minivan line private, because said early, it tells customers to wait. If you called your main product a minivan tomorrow, what would pay the bills while customers waited?

Invest Like the Best, EP.491, Michael Moritz youtube.com/watch?v=mlWX7Rlt… colossus.com/episode/the-out…

Over a million robotaxis by 2020, and the exit clause on the same stage. You've seen the clip, or at least the number. At Tesla's 2019 Autonomy Day, what left the room was one figure: over a million robotaxis on the road the following year. The promise itself is softer than the headline. In that sentence he hedges a year into maybe fifteen months, then snaps back to next year for sure. At the same event he narrows the robotaxi plan to places that allow it: "not in all jurisdictions because we won't have regulatory approval everywhere". Within two weeks Tesla raised about $2.7 billion in stock and convertible notes, and the 8-K announcing the offering talks about the balance sheet and general corporate purposes. Robotaxis aren't in it. Seven years on, Tesla still doesn't publish a robotaxi fleet count, and the cars people track through Texas registrations number in the hundreds. Last week it lined up $30 billion in credit, and the headlines said Cybercab again. This is my reading; he never said it. The million was for the room, and the clause was the forecast. Flip the sentence and it says robotaxis wherever a regulator says yes, whenever that turns out to be. A manager who repeats a vendor's date to her boss without the condition usually ends up owning it when the condition arrives. Whose date are you repeating this week without the condition?