Notes from Gokul Rajaram — Oct 16, 2024

Hot AI startups:

• Foundational models have converged around 4-5 companies. There is no more space left for foundational models. More consumer companies are building on top of the foundational models.
• Investors focus on the application layer - horizontal apps, especially new ones, won't succeed.
• For example, customer service AI or software AI startups emerge, but there are 5-10-year-old customer service software companies that have developed customer service AI, generating more revenue than all the above startups combined in a year. They hold more data and have more market share.
• There are enough horizontal incumbents, and it's so easy for them to verticalize their software.
• Verticalization of software means to go niche, delve deep, and own end to end.

Starting up vs FAANG:

• Don't join a very early-stage company - you're taking the same risk as the founder - if it's really interesting, go build it yourself - I don't think the risk-reward is worth it.
• When to build a startup:
  • Have a specific idea that consumes your thoughts deeply.
  • Startups started as a toy project.
  • Google - Stanford research.
  • Square - Jim and Jack had payment issues.
  • DoorDash - Stanford class project.
  • You can't engineer a company.
  • Don't try to be a founder.
  • Starting what you read on TechCrunch, what other people are getting funded on - bad advice.
  • No matter whether there's a recession or the first employee quits, you've got to believe in it.
• Two things - how is the company doing, how is your career doing (earnings, growing as a person, growing as a leader).
• FAANGs offer attractive packages of $1M to $2M in RSU, with immediate liquidity. You can gain credentials and build a network. If you're already there, consider joining a mid-scale company.
• 4 years each - 10 shots. A smart person will hit the jackpot in a mid-sized company, in terms of the role they own and the growth of the company. Don't make the same mistake of joining a company with the same pattern as a failed company.
• Mid-sized is 500 - 1000 people.

Why Titles Don't Matter:

• Founders should resist title inflation.
• It's not the title, it's the scope and impact.
• The counterpart is that life will be harder.
• Defer titles as long as possible.
• A CEO should not have a chief of staff.