April Musings
Quick, raw thoughts
Appetite. There's more appetite than ever for people to get in on the AI trade. Public stocks are ripping. The big hindrance to stocks, mostly the wars in Iran and Israel, has taken a back seat. Instead, people are feeling the FOMO. Even large asset managers have started reaching out asking for private investment opportunities.
No view. They surprisingly don't have as concrete a view as you'd expect of someone looking to deploy a sizeable amount of capital. It's a classic invest then investigate strategy but with FOMO doing the driving. Some of them are feeling the pain from a lack of diversification. They're realizing that the opportunity cost of spending months diligencing a single company was very very high, and that the work proved fruitless as the industry shifted to the next thing.
Entrepreneurship. Plenty of frontier lab employees and academics are exploring entrepreneurship. Capital has been unlocked but hasn't spread across the industry, and definitely hasn't spread into academia nor into the pockets of frontier lab employees who joined late. They now want to start their own companies. Look at Engramme and their supposed $1B valuation.
Borrowed ideas. Original ideas are hard to come by, or hard to tell apart. Even if you're the one to come up with it, someone else can scoop it up and run with it. People rarely give credit to where they heard an idea. As Naval Ravikant has said, it's yours now. In my recent pitch calls, I've heard several founders quote ideas verbatim from a few viral talks I've watched. If I didn't know better, I would've thought it was so deep.
Riding waves. Along these lines, it's easy to ride the wave and position yourself as having built in an area that has just recently been shown to be promising. Take Jim Fan's talk on world action models. VLAs are dated now, and WAMs are making a comeback. Now every company positions themselves as having expertise in this space. It doesn't count if you've had your finger in 4-5 pies, 4 end up going bust, and one catches a new wave.
The playbook. But the way to make money has gotten a bit easier. This is what I've observed. If you're in research or academia, publish a few papers that show early promise. Use those as the basis of your research prowess. Then sell investors on a new breakthrough that will be the foundation of your technology. You can give lead investors a hint at what this is. For everyone else, tell them it's proprietary and play coy. Leak a story to Bloomberg, Forbes, or The Information that you're raising at a ridiculous valuation. People will flock to the round. Profit.
xAI. xAI might be in trouble and the Anthropic cloud deal is the first sign. Every cofounder has left. Tribal knowledge is gone. From an employee who recently joined: No one knows how to spin up a cluster for a pretraining run. It's no surprise they struck a cloud deal with Anthropic to sell compute. SpaceX's IPO might be highly anticipated but xAI by every measure was a failed startup. People joining xAI in 2026 are realizing this. Piecemealing technology is Elon's greatest skill, look at SolarCity and the Tesla Powerwall. The same thing is happening now with Cursor and Anthropic. If he can't control DeepMind or OpenAI, Anthropic will have to do.