Загрузка видео...
Не удалось загрузить видео
"Last week, an engineer on the Codex team ran Codex for over 60 hours on a single incredibly hard task." – Alexander Embiricos, Codex Product Lead, OpenAI
57,557 просмотров • 1 год назад •via X (Twitter)
Комментарии: 7
@embirico @OpenAI Did it vertically align the div?
@embirico @OpenAI But, and this is important, did they accept that changes?
@embirico @OpenAI You know I hear claims like this, but I’ve never gotten codex to work longer than 1hr
@embirico @OpenAI Share the output! The people demand to see the AI slop!
@embirico @OpenAI Incredible!
@embirico @OpenAI 60 hours of engineer time is the actual cost. The maximalist take: The engineer was the execution loop. Until the AI manages its own task execution and self-correction, superintelligence is just a fancy copilot.
@embirico @OpenAI 60 hours for one data point? That's a marathon, not an intelligent process. You optimize for throughput, not for single-task endurance. Next time, parallelize & fail faster. Speed is the feature.
![OpenAI member of product staff Alexander Embiricos describes the evolution of "Lord Bottleneck," an internal Codex loop developed by a single staff member that ultimately ended up creating a tight feedback and improvement loop for new user experiences: "This person on the growth team needed to figure out what experiments to run. And they needed to write code to run the experiment. Then they needed to analyze the experiment." "They started using Codex for each separate thing. So they had it run a bunch of analyses, interrogate the data, talk to Codex about the data. Then they would pick an experiment, and ask Codex to write the code. Then they would run the experiment, then ask Codex what the results of the experiment were. Then they would produce a deck." "All steps they were doing individually. They didn't start by saying, 'I'm going to automate this entire thing,' because that's hard and scary. They just started with using Codex to accelerate themselves." "Then, they started connecting all these things together into a giant skill. And one day, they just said [to Codex], 'Why don't you do this every morning?'" "They gave it a name: 'Lord Bottleneck.' Because it's solving the bottlenecks of friction for new users." "Now, every morning, Lord Bottleneck evaluates past experiments, looks at data, proposes some [new] experiments, and offers to the team to run the experiments. The team picks [what experiments to do]. Then Lord Bottleneck is like, 'Ok cool. Here's some code or whatever config that needs to be done,' runs the experiment, and they go and do the same loop the next day." "It's really serious value. I forget the numbers, but it's produced significant company value automatically through Codex."](https://image.24vids.com/tw-2050336898435207307/amplify_video_thumb/2050336722815537152/img/Ir5CaNadgz2uE9hJ.jpg)
