RE: LeoThread 2025-11-11 04-08

An AI investment thesis: every AI application startup is likely to be overwhelmed by rapid expansion of foundational model providers

App functionality will be absorbed into foundational models' offerings, because the large players aren't slow incumbents (the "fast startup, slow incumbent" analogy is misleading); they are simply very large. Compared with prior technologies, a massive, fast-moving wave now obsoletes new apps almost as soon as they appear, leaving almost no time to build and scale a company

There are two ways AI application startup founders can make money:

• Create a flash-in-the-pan app that generates substantial cash and bank the proceeds (estimated cashflow window ~12–18 months)

• Build a good-enough app that is acquired by a major player for meaningful equity

The situation is highly unstable—outcomes could range from collapse to explosive growth, but both make it unlikely that an AI application startup will independently become a generational supercompany; baseline odds were already low

The best odds come from finding a niche in a highly specialized field with extremely unique, specific data barriers, ideally tied to physical-world (hardware or other real-world) data rather than purely software or finance