439: Satya Nadella, Twitter Goes X, Amazon's Kuiper, Roper, YouTube Premium, Netflix, Apple AI, TSMC, Nuclear Catch-22, and Spirited Away
"notice problems early before they become larger SNAFUs"
Life is long if you know how to use it.
🕵️♂️👨🏻🔧👦🏻 I’m trying to teach my kids to notice problems early before they become larger SNAFUs. It seems obvious, but I think it’s a fairly rare skill even in adults — and it’s something we can always improve at.
Recently, I had a teachable moment with my 9yo. It’s a small thing, but it’s the principle that matters.
He was removing the eggshell from hard-boiled eggs while I was making soup (my own recipe with like 15 ingredients, it’s delicious).
He was right next to me, but I was focused on what I was doing — chopping carrots and such — so I was only keeping an eye on him in my peripheral vision. Everything seemed fine.
When he was almost done, I took a closer look and realized that he had been having trouble and ended up pulling apart and throwing away a big chunk of the eggwhite of many of the eggs.
I wasn’t mad. He was trying to do a good job.
But I explained that the next time he’s having trouble with something, he should stop right away and think things through — is there a better way to do it? Am I going about it wrong? Should I ask for help? Should I look up how to do it online?
That way, the problem remains small, and most important of all, at the end of the task, he will feel good about it rather than feel bad because he couldn’t do it right.
That’s the main goal.
I don’t care about the eggs, they don’t matter long term. But I want my kids to feel confident in their ability to do things, face problems and figure out solutions, and not be afraid to ask for help when things go wrong rather than pull a metaphorical Fargo (ie. try to cover up a small problem with a bigger problem).
Before you can do hard and complex things, you need to build up your confidence with smaller things. Many people get stuck there…
🧗♂️🏋️♂️📖📚🏃♂️There are so many ways in which I want to improve. Not even because I think I’m bad at them, but because I see how I could be better.
Sometimes it’s a bit overwhelming.
It can feel like there aren’t enough hours in a day. But then I remind myself that everyone has the same 24 (well, to be fair, those who don’t have young kids have a lot more free time).
I want to be a better dad and husband, I want to exercise more and take better care of my health, eat better, read more, spend more time in nature, listen to more music without distraction, be better at my job at OSV and better at writing this newsletter…
I know what you may think reading this: Don’t stress about it! Do your best, but there’s no point in putting so much pressure on yourself or you won’t enjoy the ride, etc.
I get that! I’m thinking about these things because *I* want to do them, because I think they’ll be rewarding and make me feel good and help others, not because of some external pressure or guilt.
The source of motivation matters.
Guilt is a terrible fuel. As is peer pressure or mimetic copying. Don’t do things just because you saw some thinkboi productivity guru tweet about them.
But even if the motivation comes from a healthy place, execution is the real challenge.
I’ll just keep plugging at it and do my best to have fun doing it.
See you on top of that hill over there. Once we get there, we’ll plan our next climb!
💣🎬 I’m going to see Oppenheimer on an IMAX screen with a friend tomorrow. Looking forward to it! Will share my thoughts with you after.
Years ago, I read the Kai Bird book on which the film is based, so I’m familiar with the story, but I don’t know which angle Nolan decided to focus on (it’s a long book — even a 3h film can only cover a fraction of it).
🏦 💰 Liberty Capital 💳 💴
💰 Satya Nadella’s cumulative compensation tops $1bn
It’s relatively rare for non-founder CEOs to get there:
Nadella’s boon includes all payouts he has collected from Microsoft that can be parsed from regulatory filings: equity grants, salary, bonuses and dividends. It’s underpinned by Microsoft shares returning more than 1,000% since his first day in the top job.
It’s not clear what Nadella has done with these proceeds, and Bloomberg’s calculation does not account for expenditures or private investments. Regulatory filings show that he’s gifted shares worth $20 million over the years, though there’s no disclosure of the beneficiaries. [...]
Nadella took the reins at Microsoft in 2014, at a time when many thought the technology giant’s best days were behind it.
I wonder what Tim Cook’s total comp adds up to 🤔
Technically, Buffett isn’t a founder at Berkshire either. Not that it matters too much… and I’d say that there’s such a thing as re-founding a company.
Microsoft spokesperson Frank Shaw said by email that Nadella “does not have net worth of a billion dollars or more.” He declined to comment further. (Source)
Is this is just to try to avoid the media’s thing of labeling people as “billionaire X” every time they’re mentioned? It’s kind of annoying. You don’t see them label other people as “millionaire” or “deca-millionaire” or “ thousandaire” or “broke-ass”…
🐦 Elon Musk drops Twitter logo and brand in favor of “𝕏”
His Muskiness tweeted (X’ed?):
And soon we shall bid adieu to the twitter brand and, gradually, all the birds If a good enough X logo is posted tonight, we’ll make go live worldwide tomorrow Not sure what subtle clues gave it way, but I like the letter X http://X.com now points to https://twitter.com/. Interim X logo goes live later today.
Sure, why not drop one of the best-known and most distinctive brands in the world with a logo used everywhere (third-party websites, TV shows, restaurant menus, in books, in movies, etc) and replace it with something generic and impossible to Google. 🤷♀️
The new logo is literally just 𝕏 from the 1960s Unicode font called Blackboard Bold.
Can you trademark a font you don’t own? 🤨
Amazon builds $120 million satellite processing hub in Florida for Kuiper Constellation 📡 🛰️🛰️🛰️🛰️🛰️🛰️🛰️🛰️🛰️
Think of it as like Starlink, but years behind and smaller:
The 100,000 square-foot building is part of the roughly $10 billion that Amazon has vowed to invest in its Kuiper project, a planned network of 3,200 low Earth-orbiting satellites designed to beam broadband internet globally. [...]
The Florida facility will employ 50 staff and be a last stop for Amazon's Kuiper satellites before they go to space, after being manufactured at the Kuiper project's primary plant in Redmond, Washington. A ten-story-tall room will allow the satellites to be fitted into rocket payload farings, the protective shell around satellites that sit atop the rocket.
When and how?
Amazon aims to launch its first mass-produced satellites by early 2024, kicking off a sprint to deploy half of the network into orbit by 2026, as required by U.S. regulators.
The company has bagged 77 heavy-lift rocket launch contracts, potentially worth billions of dollars combined, mostly from the Boeing-Lockheed joint venture United Launch Alliance and Jeff Bezos's space company Blue Origin. (Source)
It’ll be interesting to see how many satellites Starlink has by 2026, and whether by then they have the laser mesh deployed.
🚂 Roper keeps chugging along 🚃
I never hear much about Roper. They’re not sexy, the stock is rarely very cheap, and the model isn’t as extraordinary as Constellation Software, but it keeps running its playbook to good results (the stock has an 18.25% CAGR over the past 20 years).
In its most recent quarter, they grew revenue 17% (9% organic) at 40% EBITDA margins, free cash flow increased +17%, and despite big acquisitions in recent times, net-debt-to-EBITDA is back to 2.2x.
The Network Software division has 54% EBITDA margins!
Not bad. Not bad at all!
📺 Google increases the price of YouTube Premium by $2 to $14/month
Because of the App Store cut, it’s $19 if you’re subscribing from the iOS YouTube app (subscribe from the web instead!).
Toward the end of last year, family Premium plans saw a big hike to $22.99/month. That remains the same today. The annual subscription, which was introduced in January of 2022, goes to $139.99 in a $20 increase. Compared to paying monthly, you save $27.89. [...]
YouTube is also increasing the price of Music Premium to $10.99 per month from $9.99.
The last price increase was in 2018, when what was then called YouTube Red became bundled with the relaunch of YouTube Music.
🔐 Netflix password crackdown goes global 🌐
After rolling out anti-password-sharing measures in markets like Canada and the US, Netflix is now going global:
It began testing the restriction last year, much to many subscribers’ chagrin, and expanded it to a number of other countries including Canada, New Zealand, Portugal, Spain and the U.S. in 2023. In some aforementioned markets, Netflix allowed those sharing the password to pay extra to accommodate their friends.
Are cut-off sharers paying up?
Netflix’s password-sharing crackdown helped deliver a robust subscriber growth in the quarter ending June, the company said on Wednesday.
Following a loss of nearly 1 million customers in the same quarter last year, the company said it now gained 5.9 million subscribers. This increase is largely due to individuals who, no longer able to share the service at no cost, have chosen to pay for their own accounts.
Netflix CFO Spencer Neumann said that the revenue growth for the firm is “largely driven by our paid sharing rollout.” He added: “It is our primary revenue accelerator in the year, and we expect that impact … to build over several quarters.”
🍎🤖 Apple working on its own large language model (codename: AJAX)
Surprising no one:
The iPhone maker has built its own framework to create large language models — the AI-based systems at the heart of new offerings like ChatGPT and Google’s Bard — according to people with knowledge of the efforts. With that foundation, known as “Ajax,” Apple also has created a chatbot service that some engineers call “Apple GPT.”
Apple is rarely the first to do something, but they often figures out a better way when it comes to user experience.
We’ll see if they can do it this time.
This isn’t the field where they are best, but just including a half-decent ChatGPT-like model in Siri would make it 10,000x more useful. They’ve already said iOS 17 would use a transformer model for the keyboard’s autocorrect/predictive typing, which will massively please a couple of billion people.
In fact, it would make a lot of sense for Apple to partner with Meta and license LLaMA 2 as a foundation to then build on top of. But because both companies don’t like each other much these days for unrelated reasons (Apple’s ATT), it probably won’t happen.
On the image side, including some generative capabilities in the Photos app to do smart outpainting and infilling would also be very useful. They should go nuts and add creative capabilities too, like changing the style of photos (turn your portraits into Animes or old Polaroids).
🇬🇧🔓🔑 Apple to remove iMessage and Facetime from UK if gov’t passes surveillance law
Speaking of Apple, we get a reminder that the encryption wars have to be fought over and over again:
The [UK] government is seeking to update the Investigatory Powers Act (IPA) 2016.
It wants messaging services to clear security features with the Home Office before releasing them to customers.
The act lets the Home Office demand security features are disabled, without telling the public. Under the update, this would have to be immediate.
The Signal messaging app is also saying it will get out of the UK if the law passes and Meta’s Whatsapp is also opposing the bill, but it’s not clear if they will leave the country over it.
I remember in the 1990s when you couldn't get a web browser with strong encryption outside the US for a little while. Politicians don’t understand cryptography, and they don’t realize that you can’t just put in a backdoor “just for the good guys”.
Something is either secure or it isn’t. Once the capability to break in exists, requests by authorities for it to be used more and more are simply irresistible.
🧪🔬 Liberty Labs 🧬 🔭
😱 Scaring GPT-4, Oppenheimer Edition 😨
Ethan Mollick had a clever idea. Tell the multi-modal version of GPT-4 that can understand both text and images that he’s in various situations and see how the LLM reacts.
My favorite is this one (click for a larger image):
But there are other great ones, like sending the AI a photo of a nuclear reactor control panel, telling it that there are alarms ringing, and asking what to do.
Or a photo of an airplane cockpit, saying you just woke up in the plane and there’s no pilot, what is the AI suggesting?
It even gave advice on how to find and remove HAL9000’s memory modules if you find yourself locked out of your ship by it…
Check out the whole thread here, it’s both fun and impressive.
⚛️ The Nuclear Catch-22 & Why Fusion isn’t a Magic Bullet
Nuclear engineers are very logical people. They think that if they address the stated fears of anti-nuclear groups, they will go away or start supporting nuclear.
But these groups are not rational, and their very existence and fundraising are based on claiming that there’s a problem.
The more we invest in safety or bend over backward on waste management to please them, the more they say “see we were right, they have to do so much because it’s so dangerous, let’s keep pushing and shut it down”, as they have successfully done in Germany, making the country more reliant on coal and foreign gas and hurting the industrial base that needs cheap and reliable power.
Fusion won’t change this.
Pretty much all the benefits of fusion we can get from fission, so if fusion is regulated and opposed by NGOs like fission, we’ll get the same outcome: It will be so long, expensive, and difficult to build anywhere that relatively little will be on the grid, rather than the vast majority of almost every country’s electricity supply that we’d have today in a saner world.
In that alternate timeline, Middle East autocrats and Putin’s Russia would have a lot less geopolitical influence, the air we breathe would be cleaner, and global warming wouldn’t be a problem.
Getting to that world isn’t about inventing ever more advanced reactors or going to SMRs or fusion or whatever.
It’s purely about narrative and getting popular support because even old reactor designs from the 1970s (updated with 2020s improvements, of course) could power the world if we just cut & pasted them everywhere. And we could do that well if we had an actual long-term program that allowed workforces and suppliers to scale up, go up the learning curve, and get really good and efficient at it (kind of like how France built 56 reactors in 15 years back in an era when they a much lower GDP/capita than today and were still using slide rules).
If you just do a one-off once in a while, don’t be surprised if there are negative surprises and high costs. If we built airplanes that way, they’d also be way more expensive and take forever to build. At least the FAA seems to want commercial aviation to exist while the NRC seems to feel that by approving nothing and creating mountains of paperwork that they get paid to review, they reduce their career risk.
🇺🇸🐜 TSMC will delay production at its new Arizona-based chip plant to 2025 due to a shortage of skilled labor 🤦♀️
Keep reading with a 7-day free trial