Combat is on my mind this week. Companies are battling over researchers, over GPUs, over market share. It is bracingly brazen and aggressive, but that is what a paradigm shift feels like. This week we will talk through some winners (and some losers who are lying about winning).
But first, this Weekend Leverage is brought to you by the always tasteful Notion.
My best ideas always come in conversation. I’m guessing yours probably do too. However, some spark is always lost when I hastily try to write down my thoughts after the meeting wraps up. Notion’s AI Meeting Notes solves all that. It uses LLMs to record your meetings, and more importantly, integrates those conversations into your existing databases and workflows. Now when I’m researching a piece, I’ll record the conversations directly into my content schedule. It’s four tools all at once. The result is a product that saves me significant time and improves the quality of my writing. It's a glimpse of the future of work—faster, smarter, better.
MY RESEARCH
Where are we overestimating AI? During a bubble, capital is deployed on everything related to a trend. However, when the bubble pops, only a select few companies emerge as centers of profit. For now, VCs are investing like everything is going to win. Tech crossover investor Coatue—they have $50 billion in assets under management, so it’s worth paying attention to what they say—published a deck outlining their thesis. Where they get it right is informative, but what they get wrong is an insight into the entire state of the tech industry. Let me know how you liked this one, I was surprised at how positive the response has been so far!
Slack declares war. Every old-school SaaS application is going to do their best to prevent you from accessing your own data. Meanwhile, every AI application is going to do their best to get access anyway. Who wins? Who loses? The answer to this question matters because the more data access an AI system has, the more likely it is that your job is going to get automated away.
THE BIG STORIES
Lies, damned lies, and Marc Benioff’s marketing: In an interview with Bloomberg, Salesforce CEO Marc Benioff declared that “AI is doing 30% to 50% of the work at Salesforce now.” If you watch the interview, I think this sentence is 20% brain fart—he was previously talking about engineering and customer service, so that productivity gain may not be company wide. Still, that is a HUGE, GIANT thing for a public CEO to claim on the record. Benioff is known for having a complicated relationship to truth and marketing, so I wanted to investigate it further. Spoiler, it's probably bullshit. Below is a chart of Salesforce’s revenue and the number of employees over the last five years as pulled from their earnings reports.
If there was this level of efficiency gains, I would expect either of these lines to bend somewhere. This is something I’m starting to notice over and over—companies overexaggerating the impact of AI because the impact allows them to increase customer purchasing urgency.
Waymo is the most important company in America: As I’ve written about before, cars kill roughly the same amount of people in the United States as firearms do—40,000 annually. Simultaneously, Waymo has 85% fewer accidents with serious injury than human drivers. We should all be cheering for autonomous vehicles to work this well! The primary competitor in the U.S. is Tesla, which kicked off its robotaxi pilot in Austin this week with ten vehicles.
Waymo’s vehicles have Lidar, radar, and camera sensors. Their software is a mix of neural net programming and deterministic code. Tesla chose a minimalist approach with just cameras and an end-to-end neural network. This means that Waymo is safer, more functional, and can currently run commercial programs in four different U.S. cities. However, compared to Tesla, it is far more expensive to run and maintain. Tesla’s technology, to be very clear, does not currently work for safe, fully-autonomous driving. They are very, very close, but are limited by the camera-only approach.
At various points, Elon Musk has said there would be “millions” of robotaxis on the road by 2026, but it feels safe to dismiss that. More likely, the robotaxi pilot will scale to roughly 1,000 vehicles in a larger testing fleet that Musk has discussed. When you incorporate Waymo’s stated forecast, robotaxi fleets should look like this next year:
If you triangulate between industry estimates, each Waymo vehicle currently costs somewhere between $120-$150K to manufacture. Tesla’s robotaxi will (supposedly) cost around $30,000.
If true, all these data points put together means that the instant Tesla’s robotaxi is safe enough, it will be able to scale up dramatically faster than Waymo. Meanwhile, Waymo is aware of this and is frantically incorporating neural nets and focusing on decreasing costs. It’ll be a fascinating race to watch.
Amazon vs. SpaceX is a story of bundling dynamics. Starlink is another incredibly impressive project in the Musk universe. The internet service is fast, reliable, and globally available. The only other company even remotely trying to chase it is Amazon with their Kuiper project. Amazon launched some satellites this week to bring its fleet to a grand total of 54—it uh, doesn’t look so good compared to SpaceX, which is currently approaching 8,000 satellites in orbit. Take a look.
It is easy to dunk on this effort, but Amazon is staffed with some smarties. They know the difference in launch costs and in time to market. However, what feels most likely is that they are counting on folding it into the Amazon Prime bundle for consumers, or with AWS for enterprise clients. If business is the process of unbundling and rebundling services, then I’m interested to see how they position satellite internet as part of Prime. It feels funny to simultaneously get my diapers, James Bond movies, and satellite internet all from my Prime subscription, but hey, I’ll take it if it's included for free!
TASTEMAKER
28 Years Later: Good horror films fill you with dread, great ones do the same, but then make you think more about why you are feeling that way. The first act of 28 Years Later is Grade A, primo scary zombie-movie material. Fantastic and fun. Then, in the second and third act, the film is bold enough to swerve into questions of masculinity, violence, and love. My favorite new film of the year so far. If you aren’t convinced, just watch this spoiler-free trailer and try to tell me you don’t have electricity in your veins.
My most anticipated summer album: Lorde’s previous work was always enjoyable for me, not earth shattering. This album is her best ever. She embraces the synth-driven pop of her past, but now approaches it with a degree of wisdom and seriousness previously missing. I’ve listened to it three times in 24 hours—this will do very well at the Grammys this year.
My favorite book I’ve read this year—The Living Mountain: Have you ever finished a book, flipped that last page, taken a deep breath, and then immediately returned to the first page to start over? The Living Mountain did that for me. The author, Nan Shepherd, spent her entire life hiking around the Cairngorms mountains in Scotland and this book is her beautiful, meditative reflection on those decades spent in the wilderness. Most nature writing is about the pursuit of the summit or the end, Shepherd’s efforts are spent being present, embodying the nature of which she is a part. The writing is melodious, yet sparse, and had me reading passages to my wife multiple times a week. Highly, highly recommended.
Until next week, my friends. Paid subscribers to The Leverage can look forward to two paywalled essays this week containing my best and most in-depth research. You can upgrade below to get them in your inbox.
Thanks for the great compilation!