Come Build With Me in SF
The Weekend Leverage, August 2nd
Friends, last week I announced The Leverage Launch—a one night build devoted to making beautiful, weird websites. I’m excited to announce we have secured Cursor as the presenting sponsor for the event. They were really excited about my vision for a more aesthetic internet and stepped up to make my dream possible.
We will have free dinner, free drinks, free tokens, the first-ever merch for The Leverage, Dandelion Chocolate bars — because why the hell not, I’m using this sponsor money to ball out, not wimp out — and a luxurious leather-bound prize for the winner of the evening. Attendees already include a New York Times bestselling author, senior operators at some of the best AI companies in the valley, and investors with copious amounts of drip. Really, though: this is about bringing together people who want to make the internet overflow with swag and dopeness.
Please join us! Attendance is free because I love you.
When Ilya Sutvesker, the former co-founder and Chief Scientist of OpenAI wanted to rally the troops, he would have them chant out, “Feel the AGI! Feel the AGI!”
This week I felt a tickling along the back of my elbows, my neck hairs standing at attention, a robotic whisper in my ear. I didn’t fully feel the AGI, but I felt/heard/saw its coming promise. We’ll get to why in a second. First we need to talk about the stupidest website I’ve ever made.
I asked Grok if I was full of shit. Musk’s chatbot scored me somewhere in the “low-to-mid range for a tech newsletter guy,” which as an egomaniac I found deeply offensive. As an analyst I was annoyed because it wasn’t very precise. So I spent a day using ChatGPT and Claude Code building isevanfullofshit.com, a machine that ingests everything I’ve published and grades my forecasts. This is clearly dumb. But my point here is that my exercise is one that everyone everywhere is about to do with LLMs. Point it an archive of any data type and it’ll tell you, with sometimes undeserved confidence, what the data means. This will happen to Slack, text messages to your ex, everything you’ve ever said at work. I call this “semantic surveillance” and I dug into who’s already selling it. Read here.
OpenAI’s next model solved open questions in mathematics using $2000 of API tokens. Late Friday evening, my phone started blowing up with texts of “did you see this?” OpenAI used their new model, code name Astra, to solve 10 different outstanding math problems in areas like “arithmetic circuit complexity.” Now I have absolutely no idea what that means, but math professors on X are calling it “a big deal.” Capitalist pig that I am, I immediately thought about the costs. They did these 10 problems with $2,000 of tokens.
I would caution against saying that this will soon happen for everything. Math has verifiable proofs, which means there are victory-conditions for these models that are verifiable. Still, for problems with definite answers, there is a decent chance they may all end up being solved problems over the next few years. (Isn’t that a crazy sentence to read?) Lots could happen, but still, these models keep getting better, faster.
Nvidia’s revenue polycule keeps getting messier. After two years of monastic silence, Ilya Sutskever’s Safe Superintelligence announced a partnership with Nvidia that includes a reported $5 billion investment. The company says its research finally reached the point where it deserves scale, and Ilya posted “Time to scale that SSI” on X. This is the same man who has declared multiple times that “the scaling era is over” because the previous research paradigms wouldn’t get us to AGI. So, him taking this money now likely means he believes he has a research direction worth scaling once more. SSI still has no product, no revenue, and a few dozen employees, which puts it at roughly $200 million of capital per head. Not bad!
Nvidia keeps doing this very weird circular revenue game where they invest into their customers, who are all competing with each other. I tried to map them all, but there are so many of these deals: a reported $250 AI, $10 billion into Anthropic, checks into xAI, Mistral, Reflection, and Thinking Machines on one side, and stakes in CoreWeave, Nebius, and (soon) IREN on the other.
Keep in mind that neoclouds like CoreWeave are buying Nvidia hardware to rent to the labs Nvidia also funds. When the same firm is the financier, the supplier, and the appraiser, my inner security and exchange commission starts squirming.
LinkedIn attempts to deslopify. LinkedIn shipped a “Seems like AI slop” button this week. The company is simultaneously killing its own “enhance your post” AI writing feature, a rare case of a platform having to publicly say “whoopsie.” The platform is riddled with it. AI detector Pangram (disclosure, former sponsor) found that 41% of long-form LinkedIn posts are entirely AI-generated. This is by far the highest amount of the major internet platforms.
This actually places the company in an incredibly nasty bind. If you enforce this you vaporize two-fifths of your content supply overnight. Don’t forget that people LOVE AI slop. Their revealed preferences, over and over again, is that they are willing to read and pay for it. If you try to keep your platform AI free, that means your feed shrinks, sessions shorten, ad inventory evaporates, and then your ass and the curb are soon to meet. But if you do nothing and the ratio keeps climbing, you’ll likely rot the platform from within. It is a variation on one of my theories I’m best known for, “The Content Double Bind.” Platforms have no choice but to allow the most socially corrosive material because it is the most engaging type of content. If they don’t allow it, their competitors will. I do not envy the LinkedIn product team trying to solve this.
The only good interview of The Odyssey press tour. A humanities Phd who actually understands Greek psychology and mythology pushes Nolan on his Christian modernization of the classic tale. Watch.
Mr Rogers was one-of-a-kind. I have started to watch the classic children’s program with my daughter. It is peaceful, meandering viewing, something that stands in stark contrast with modern editing. I was reading about his life and found this speech where Rogers convinced Congress to give $20M to public television. His tone and passion gave me goosebumps. Watch.
The Matic Robot is the greatest electronics device in my life. I thought the hype would be overdone, but this little robot vacuum is INCREDIBLE. It is somehow making a dent into the mess that the grime tornado that is my toddler makes on daily basis. Really great technology paired with a delightful consumer experience. Seriously, go get one. Buy.
Go and be kind this week,
Evan
Sponsorships
We are now accepting sponsors for Q1 2027 (crazy!) Please reach out if you want to reach my audience next year.