“There is no analogy for what we are making. You cannot copy any existing thing,” Lenny on Saturday says Nick Turley, the head of chat in Openai in Rachitsky’s weekly tech podcast. And it is exactly how he thinks that employees should work. According to Turley, there is not only technical skills or large-compani experiences held in Openai. It is about the manufacture of products from scratch, opposed the urge to follow the playbook of Big Tech, and learns from the user reaction in real time. “It is very important in this place to complete each scenario from scratches,” they say.

Turley reported that Openai cannot customize only the facilities developed by technical giants such as Instagram, Google, or companies known for productivity devices. “You can learn from everywhere, but you have to do it with scratches,” he said. “This is why this characteristic goes to make someone effective in Openai, and this is something for which we test.”

The conversation also touched Openai’s approach to launch the products. The company unveiled GPT-5 last Thursday, received mixed reviews. Turley said that the openIE intentionally releases features before it is completely refined.

He said, “We get a lot of nonsense for the model selector,” he said, referring to a now-set dropdown in the chat, which users select between Openai models. They believe that it is better to release a rough version to gather the reaction of the real world, rather than that until a product meets the internal quality bar.

“The reason really is that you are going to brighten the wrong things in this place,” said Turley. “You will not know what is polish till your ship”, a theory he believes that when the product behavior emerges in an unexpected manner is particularly relevant.

The GPT-5 launch shows this approach in action. The controversial model picker has been replaced with “real -time router” that automatically chooses the best model for the user’s request. “Earlier, you had to deal with model picker in chat,” COO Brad Lichap of Openi told Big Technology on Friday. “You had to choose a model that you wanted to use for a given work, sometimes you choose a thinking model, sometimes you don’t. And it was, I think, a misleading experience for users.”

Nevertheless, pleasing everyone is a challenge. After the release of GPT-5, some users complained that a very loving old model had disappeared. Openai CEO Sam Altman later confirmed that it would be brought back.