Commentary on X
@emollick the constant tweaking shows how models are evolving it highlights a balance between deep iteration and efficiency which is key for practical use in workflows
8:11 AM · Jul 16, 2026@emollick yes it rethinks a lot, in a smart way. i had a solid first impression.
9:21 AM · Jul 16, 2026@scaling01 (like since it's slow you have the feeling it's thinking a lot but in reality it's not)
8:53 AM · Jul 16, 2026View all
@emollick All that capability and it still can't tell when a job's finished.
8:45 AM · Jul 16, 2026@scaling01 @eliebakouch has been overthinker since K2.5
8:59 AM · Jul 16, 2026Kimi K3 seems really good, closest to the frontier yet, but also wow does the model/harness love to loop back over and over again over tasks tweaking and changing things at max level. Anyhow, examples incoming, when they finish.
7:59 AM · Jul 16, 2026@scaling01 isn't it just an artefact of it being slow?
8:52 AM · Jul 16, 2026