[granitepail]

"Just" Kimi K2.5 with RL—people really misunderstand how difficult it is to achieve these reults with RL. Cursor's research team is highly respected within the industry, and what they've done is quite impressive.

Before people go jumping to conclusions about model theft, it's worth considering the possibility that they did reach an agreement with Moonshot which their researchers were not aware of. That would certainly explain the deleted tweets. Until Moonshot makes an official statement, I'm not particularly concerned.

[halJordan]

The amount of angst people feel the need to have against ai is incredible. We all seemed to want open weights, but it's time to take offense when open weights are used as intended?

[granitepail]

Hey, what do you know—Moonshot came out and explained that Cursor was operating within the licensing agreement. https://x.com/kimi_moonshot/status/2035074972943831491

[Aurornis]

Earlier their employees were Tweeting that Cursor did not have a license ( https://pbs.twimg.com/media/HD2Ky9jW4AAAe0Y?format=jpg&name=... ) so Moonshot employees are partially to blame for the outrage.

[ajitid]

Could you explain how much improvement RL+fine tuning has given to Composer 2.0 over Kimi K2.5? I don't fully grasp the work Cursor model has done here and why it is difficult to achieve these results with RL.

[jeany2s]

You should realize the number/comparison they released is based on CursorBench... I don't have to emphasize more how sus this kind of self-defined closed-source benchmark can be