Глава Codex Эндрю Амброзино рассказал, как внутри OpenAI проверяют новые модели.
Глава Codex Эндрю Амброзино рассказал, как внутри OpenAI проверяют новые модели..
- 01.Глава Codex Эндрю Амброзино рассказал, как внутри OpenAI проверяют новые модели..
Codex head Andrew Ambrosino [talked]() how new models are tested internally at OpenAI. Every time a new model comes out, he gives it the same task - to completely rewrite the desktop Codex in Rust and GPUI. Not to make a demo, but to build a full-fledged native application that works the same on different platforms. According to him, the difficulty here is not only in the code, but also in coordination. The Codex team changes the product so quickly that the model literally does not have time to work with the committed version. The idea is that one strong enough model can simultaneously support a Mac version on AppKit, a Windows version on WinUI 3, and a separate full-fledged version on Rust. “We're not there yet. But it’s much closer.”
Related Intelligence
All stories →OpenAI 推出 GPT-6.1 Terra,新模型成本远低于预期。
OpenAI 推出 GPT-6.1 Terra,新模型成本远低于预期。API 使用费用:> 每百万输入令牌 1 美元。> 每百万输出令牌 5 美元。> 每百万缓存输入令牌 0.05 美元。因此,GPT-6.1…
OpenAI 如何将其编码工具转变为 dots,一个始终在线的个人助手
负责 dots 的 OpenAI 高管 Alexander Embiricos 从不想与别人的 AI 代理开会。
It's happening...
One of the most controversial companies in history just filed to go public. And yes, we're talking about Anthropic. The same company that says there's a greater than 10% chance AI kills all of humanity is also simultaneously one of the best built businesses of all time. And in…