This is quite interesting. When the AI lab announced their results: "Look, this fully proves our breakthrough." But when the outcome didn't meet expectations? They immediately changed tune: "Benchmark tests can't really measure true intelligence levels."

Selective belief in data—this trick is tried and true in the tech industry. The problem is, you can't use benchmarks as a success metric and then claim the benchmarks are invalid when you fail. Either benchmarks are meaningful, or you shouldn't mention them at all.

This attitude reflects a phenomenon in the industry: when the data favors you, it's ironclad proof; when the data is unfavorable, you start questioning the validity of the test itself. Truly capable projects should have a clear understanding of their results—celebrate wins modestly, and don't blame others when you lose.