Has anyone seen any solid proof that Claude 4 can actually write code continuously for nearly seven hours? I'm curious whether this is genuine progress or just clever wording. By 'real progress,' I mean can the model truly decide to keep working autonomously for hours without someone nudging it to continue?
2 Answers
That sounds impressive, but I'm wondering what it even means for it to write for seven hours straight without testing. Who codes like that without constantly checking their work?
Yeah, I've heard reports about Claude 4! Apparently, it maintained focus on a complicated open-source project for almost seven hours during tests at Rakuten. This is a huge step up from past AIs that could only handle a few minutes of work. It seems like it's evolving into more of a collaborator rather than just a quick-response tool. There's a really detailed article about it if you're interested!
Maybe it includes writing the code, testing it, and fixing bugs all at once? That would make more sense.