My experience with Anthropic and OpenAI is they’re not super interested in experience and don’t take internal references very seriously. Most of who I know that was hired were fairly junior folks and they have an early weed out that’s a fairly rudimentary but very specific Python programming quiz typically administered by very junior (like 1-2 years out of school) - even when interviewing extremely senior and experienced people of some substantial success and renown. This isn’t uncommon - meta and others do this too. But the programming quiz at Anthropic is sudden death and the first round, and the people administering it are looking for a very specific implementation that if you don’t see it immediately they just Gen Z stare and don’t discuss etc. It’s one of the more amateur selection processes designed with an extreme bias against more senior folks (frankly it felt unintentional just naive). (Meta etc scale the programming weight to seniority and the administrators scale as well - asking for depth of understanding of concepts as seniority grows with the expectation experience brings more to the table than syntactic knowledge).

So getting an internal reference and being highly qualified for something they need done isn’t enough. You need to also make it past the 20 years old gate keepers and their amateur hour hiring process.