the new type of startup

most AI startups will be dead in 6 months

the AI labs have been on a killing spree lately. from anthropic launching claude tags in slack, google notebook lm short form info content, I can go on with the examples. so much startups believe they have the moat, but do they really?

we both know wrappers are inevitable to fail, but even startups with the most creative use of any of the major AI labs API's are never truly safe. the application layer will eventually be dead.

there are two options you have:

be apart of the application layer at these ai labs

OR

build in data

what do I mean?

major llm's still struggle with domain specific tasks, not to mention frontier robotics labs still need better quality data to actually be effective. the bottleneck was never the llm, the quality of data just decreased.

kled.ai forget them raising a crazy amount of vc. consumer data marketplace makes sense; llm's need better data in niche areas + better resonance with consumer, robotic companies still need better real world data that is abstract, and the value prop is clear: you get paid for your data

mercor? getting experts to work with ai labs to get even more domain knowledge.

the data market is a zero sum game. you can build in edtech (extracting high quality math problems from students so llm's can better explain concepts) agricultural, the future of ai integration in farming needs better data, I can go on

data marketplaces are the most longevity based startup niche you can break into if given a clear value prop to the consumer giving this data.

the hard part is getting enterprise consumers (chicken-and-egg problem) especially for ai labs.

regardless the longevity argument stands true, data marketplaces are still super early to break into. it is never to late

note: you can 100 percent build a saas rn for the next few months and definitely scale it to 10k (maybe even 100k) mrr, but if you care about longevity maybe pivot. i also believe consumer apps don't apply to this