Small Teams Are Beating the Big Labs at Weird Niche AI
The frontier models get the press. The profitable AI is happening in boring verticals with three employees.
The headline AI is always the biggest, the smartest, the most expensive. The useful AI is often a three-person team solving a problem a giant would consider beneath it.
Example: a shop that built a model to read messy handwriting on freight forms. No lab is racing them. They charge freight companies real money and have no competitors worth mentioning.
Big labs optimize for benchmarks and demos. Niche teams optimize for a single ugly workflow that someone will pay to never do again. Different goals, different winners.
The risk for the small teams is the platform shift. When a giant decides the boring vertical is worth a feature, the moat gets thin. But by then the small team usually owns the customer relationship, which is the real moat.
If you are starting an AI company, do not try to beat the frontier. Go where the frontier is bored.