I have consumed enough NyQuil to fill the Nile. The Amazon rainforest trembles at the mere mention of my child’s nose and the tissues it requires. My carpet has absorbed three bottles of Oxiclean Carpet Stain Remover, Dog Vomit™ edition. My wife has had to ask Google, ChatGPT, Bing, and Deepseek variations on the question, “Breast milk, cold medication, safe??”
All of which is to say The Leverage was on sick leave this week. I apologize for not publishing any deep dives. The whole family, dog included, came down with something diabolical. But this Friday morning, I have consumed 300 milligrams of caffeine and will, come hell or high fever, deliver The Weekend Leverage to you.
Tech is playing a game of chicken with AGI. Every firm is optimizing toward that outcome, from Meta’s $100 million signing bonuses for researchers, Amazon and Microsoft cutting headcount to free up cash for AI infrastructure, and VCs pouring more and more capital into the category. If this doesn’t work, the resulting market implosion will make the 1999 bubble look tiny.
But first, I’m proud to announce the Weekend Leverage’s newest sponsor, Notion.
I have found that the most important data is the kind that isn’t neatly compiled. It's in the conversations that I have with sources, a random PDF, or some hastily scrawled notes on a lost document. That’s why I was so interested in Notion’s newest AI offering: research mode. It turns my natural language query into a search of all the data contained in Notion and across the internet. 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
Why aren’t AI agents working yet? Despite Salesforce’s insistence on “Agentforce” you shouldn’t be fooled—AI agents aren’t all that successful. They can’t fully replace anyone’s job. Still, you shouldn’t fall for the temptation of calling it vaporware —the barriers stopping agents’ progress are beginning to topple. I’ve invented a new framework I call the “Legibility Index” which allows you to forecast when and how agents will be ready to take over portions of your job. More here.
THE BIG STORIES
Short form video has taken over the world. At Cannes, the ad conference in the French Riviera, YouTube CEO Neal Mohan finally provided some data on the explosion of short form video on the platform. I mean, just look at this:
That is 200 billion views a day. The median length of a short video is about 30 seconds, meaning that 100,000,000,000 hours of short form videos are being watched, or 11,415,525.11 years of our lives are being sacrificed to the attention gods each day. This is just YouTube! When you add in Reels, TikTok, Douyin (the version of TikTok popular in China), we are talking about a wholesale hijacking of human attention. When I say hijacking I mean that literally. A few weeks ago, I attended a workshop from prominent YouTubers on how to make short form videos that would perform well. They spent most of the hour-long lecture focusing on the first five seconds of each video. In the first second, they say, you need a “scroll stopper,” visual action that strokes some part of our lizard brain. It could be something gross or sensual, maybe it’s perfectly balanced like a Wes Anderson shot, or maybe there is lots of motion. From there you needed a “hook,” some form of a question or interest that doesn’t get resolved until the last second of the video. Both the scroll stopper and the hook happen in the first five seconds.
This is a finely tuned algorithm, designed to hijack the quirks of human evolutionary psychology.
It is easy to reflexively moralize against this stuff. It feels icky that creators are rewarded for brain hacking instead of quality. I feel this way a lot of the time. Still, you gotta acknowledge that this is what consumers want. Shoot, this is what you, fans of The Leverage, want! I get lots of emails from readers asking me to do videos, reels, and podcasts. Writing will always be core to what I do, but short-form is what the world is demanding now.
Everything is entertainment: A few years ago, I met with this really incredible founder who told me that livestreaming was the future of shopping. I thought he was great, and that I should find some way to join, but never followed up. This week that founder’s company, called WhatNot, got a glowing feature in Fortune forecasting that it should hit roughly $1 billion in revenue this year. That my friends, is what we call a generational wealth kinda miss. Whoops.
The company operates a shopping livestreaming platform. Think of it like QVC, but there are thousands of channels, each run by an independent merchant. They sell collectibles like FunkoPops, Pokemon cards, coins, sports memorabilia, and other such nerdy trinkets. WhatNot’s fastest growing segment is women’s fashion, so it’s starting to branch out into a broader market.
What fascinates me most is this little stat, slipped in near the middle of the piece. “WhatNot users spend 80 minutes on average on the app each day.” That is an incredible amount of time. To update the chart I shared a few weeks ago, this puts the company in the same realm as TikTok and YouTube.
We are likely at peak screentime. Americans already spend about six hours and 40 minutes a day on average on screens, which means that every service is in a state of competition with every other company for your eyeballs. Software can’t just be useful, it has to be entertaining, too. WhatNot does this by fostering community engagement with the merchants. Viewers can participate in a chat and send emojis, similar to a Twitch stream.. It is worth considering how you can make your company’s tooling fun.
Media’s internet distribution engine has officially collapsed. An ongoing story in the media industry is how Google and social media are no longer a source of site traffic. This has happened for many years, but the arrival of chatbots has dramatically accelerated the decline in the last three months. We finally have some data on this from CloudFlare CEO, who shared the following at Cannes:
Note that the scale is logarithmic! The increase in crawlers to visitor ratio means that Chatbots are sending very, very minimal amounts of traffic to publishers. Services like ChatGPT will crawl hundreds of websites, compile their work, and then maybe surface 5-10 sources. And then, users aren’t really clicking the sources anyway! It is hard to overstate how devastating this is for the internet economy as structured. Bad, bad, bad. Content publishers are going to have to reinvent their business models and staffing from the ground up.
Amazon and Microsoft are cutting headcount. From Amazon CEO Andy Jassy, “In the next few years, we expect that [AI] will reduce our total corporate workforce as we get efficiency gains from using AI extensively across the company.” And then this Bloomberg headline, “Microsoft Planning Thousands More Job Cuts Aimed at Salespeople.”
AI will make work more efficient, but in my opinion, it hasn’t happened terribly broadly yet. Instead, read these as “we believe AI is coming tomorrow and are spending like it is here today—work hard or lose your job.” CEO’s are putting the fear of God into their workforce. The era of rest and vest is likely over.
VISUAL SIGNAL
I worry for the students. Suffering is kinda the point of school. Learning to sit with difficult problems, to decipher challenging texts and assignments, are the the skills you are supposed to gain. ChatGPT changes that. It smooths away the friction, dilutes the experience into prompt, output, repeat.
All of my friends in academia have been panicking over the decrease in student attentiveness and quality of essays. Their sense is that shortform video is reducing students’ capacity to pay attention, and that ChatGPT is giving them the answers, but they didn’t have any data to back it up. Their guess is that students are blatantly cheating because they just view their degree as a box to check before they get on to their job. It is incredibly difficult to prosecute a student for AI cheating, so they think that 95% of it goes unreported.
We finally have hard data to back up that anecdotal evidence. The Guardian sent 155 universities Freedom of Information requests about student cheating rates. They found that as claims of plagiarism were falling, AI-related academic dishonesty was increasing.
What feels more likely—AI cheating is way lower than plagiarism? Or that students got access to tools that aren’t being detected by current cheating detection systems? I think the latter.
This is not an “AI melts our brains” argument. I know, through personal experience, how much better of a researcher and thinker the proper AI tools and strategies can make you. However, if this newsletter edition has any theme, it is that humans default to easy mode. We want reels not essays. We want prompts not homework. It takes a unique individual to lean into suffering and our cultural response to AI has not encouraged that uniqueness.
This trend is only going to increase as AI gets smarter. It is something that I’m thinking about a lot with my own kid—how do I prepare her for the AI world to come?
TASTEMAKER
Why VC funds keep getting bigger: A big story in tech is the rise of mega-funds. It’s an analysis I keep picking at, trying to understand which fund structures will win, and which will best support founders. This interview with Bucky Moore on why he joined a mega-fund, conducted by the wonderful Molly O’Shea, helped answer some crucial questions for me. The short answer is that these funds believe there are multiple trillion-dollar companies being built right now, and you need a fund that scales with them in their ambition. Worth the time to watch—I found his answers more honest and thoughtful than most of the people pursuing this strategy.
Meta CFO’s first ever podcast appearance: There have been multiple articles this week on how Zuck&Co. are spending tens of billions on hiring the best AI talent possible. $100 million signing bonuses? Sure. Buying out venture funds to get the partners to join? Why not? Reading these headlines, I kept thinking about the poor CFO trying to handle the fallout. This is her first pod interview and there is a section where she talks about Meta’s tolerance for higher-risk bets than most corporations. The section of talent development was excellent too.
Songs to teach your kid to rock: I am trying to make sure that the next 10 years of my life aren’t spent listening to Disney soundtracks on repeat. In pursuit of that, I’ve been playing and singing different genres of music to my daughter every week. This week is rock education and I prepared a playlist for the occasion. The playlist has no explicit lyrics, is composed of all up-tempo energetic songs to hold her attention, and it samples everything from classic rock, to Australian punk to midwestern emo. I thought some of you may be in the same boat. Enjoy! It has some of my favorite guitar-fueled bangers. Track 1 in particular, rips.
Tasteful companies assemble
With Notion joining as our newest sponsor, The Leverage’s brand vision is becoming clearer. We help companies connect with over 35,000 tasteful, discerning decision-makers in Silicon Valley. If your company builds products aligned with this ethos, please reach out—we have a few sponsorship slots remaining for late summer and fall. Just email team@gettheleverage.com.
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.
"It feels icky that creators are rewarded for brain hacking instead of quality. I feel this way a lot of the time. Still, you gotta acknowledge that this is what consumers want." That's because it *is* icky, and what you do next after acknowledging it's what people want matters *a lot*.
Just because people want it doesn't mean you (or others) should give it to to them, it depends on your goals. Is profit pretty much all that matters to you? Or "engagement", for whatever reason? Great, yep, do it. But not all mediums are equal for every message, every type of content, every underlying goal. The medium, if not "being" the message, at least significantly affects it and its meaning.
If what you want to do as a content creator is communicate meaningful and actionable insights on sometimes complex business topics that others are often missing, even *if* you were able to compress that into a 60 second video, you have to wonder if the people engaging with that would actually take it in, or would they lose it immediately as the next 60 second video demands their attention. Would they benefit from it as much as an in-depth read? Almost certainly not.
People may *want* the 60s video, but people want drugs, excess sugar, gambling, and a ton of other things that are only healthy in moderation *at best*. It's a typical moralizing argument but that doesn't make it any less true. Individual agency can't be a reasonable out when you're using a platform that relies on a mechanism that intentionally subverts the psychological underpinnings of true agency. "I've drugged you with a sedative, but don't worry, you can leave any time" is literally a direct comparison, as exaggerated as that sounds.