How can distilled opus become better than original? There are numbers of reports including anthropic that kimi team was participating in fraudulent activities
Agentic reasoning and tool use
Coding and data analysis
Computer-use agent development
Computer vision
Moonshot (Kimi models) employed hundreds of fraudulent accounts spanning multiple access pathways. Varied account types made the campaign harder to detect as a coordinated operation. We attributed the campaign through request metadata, which matched the public profiles of senior Moonshot staff. In a later phase, Moonshot used a more targeted approach, attempting to extract and reconstruct Claude’s reasoning traces.
And when you here unsubstantiated rumours* that say Anthropic has been sending exchanges to say Alibaba's Qwen, will you als oconclude the same about the entire US AI industry?
I feel distilling something as big as Opus would require many many more samples, but I dont really know much about this subject