I completely ignored Anthropic’s advice and wrote a more elaborate test prompt based on a use case I’m familiar with and therefore can audit the agent’s code quality. In 2021, I wrote a script to scrape YouTube video metadata from videos on a given channel using YouTube’s Data API, but the API is poorly and counterintuitively documented and my Python scripts aren’t great. I subscribe to the SiIvagunner YouTube account which, as a part of the channel’s gimmick (musical swaps with different melodies than the ones expected), posts hundreds of videos per month with nondescript thumbnails and titles, making it nonobvious which videos are the best other than the view counts. The video metadata could be used to surface good videos I missed, so I had a fun idea to test Opus 4.5:
把强模型的输出喂给弱模型，弱模型能快速获得类似能力——这个逻辑本身成立，Lambert 没有否认。但他指出了一个没人说清楚的问题：蒸馏的天花板到底在哪里，取决于你想要的是什么类型的能力。
，这一点在safew官方版本下载中也有详细论述
Most digital images intended for viewing are generally assumed to be in sRGB colour space, which is gamma-encoded. This means that a linear increase of value in colour space does not correspond to a linear increase in actual physical light intensity, instead following more of a curve. If we want to mathematically operate on colour values in a physically accurate way, we must first convert them to linear space by applying gamma decompression. After processing, gamma compression should be reapplied before display. The following C code demonstrates how to do so following the sRGB standard:
And Blue Origin responded, "Let's go! We're all in!"