Why are there so few independent eval startups?
Summary
This article argues that independent eval startups face structural problems that make them difficult to sustain. It says strong eval talent usually moves into post-training or application development, where the money and influence are greater. It also says many potential customers either can build evals themselves or are too technical for eval tooling to be an easy sell. The piece adds that large model labs can game public benchmarks, which weakens the value of standalone eval businesses. It concludes that safety evals are the main niche where an external vendor can still have a defensible role.
Classifications
industries
No industries detected
applications
Web and Content Management
AskAI Classifications
Labels
AI Software
SaaS
Developer Tools
