This is what I'd expect to see if OpenAI had made general progress on fighting sycophancy and manipulation. :/ If that's in fact what happened, OpenAI made that choice rightly.

To the other companies: it might sound like a profitable dream to have users love your models with boundless fanaticism, but it comes with a side order of news stories about induced psychosis, and maybe eventually a violent user attacking your offices after a model upgrade.

Remember, your users aren't falling in boundless love with your company brand. They're falling in boundless love with an alien that your corporate schedule says you plan to kill 6 months later. This movie doesn't end well for you.