Anthropic's original take home assignment open sourced (github.com)

Anthropic has open-sourced its original take-home assignment, allowing developers and researchers to measure their performance against the Claude Opus 4.5 language model. This initiative comes after Claude Opus 4.5 demonstrated superior capabilities, outperforming leading human benchmarks in just two hours of task engagement. Interested participants now have the chance to challenge this benchmark with unlimited time, making the competition more accessible while fostering innovation and exploration within the AI community. This release is significant for the AI/ML community as it encourages collaboration and transparency, enabling practitioners to test their skills and methodologies against advanced AI models. Key technical details highlight that Claude Opus 4.5's performance was measured in clock cycles, with the best outcomes recorded at 1487 cycles after extensive testing. Participants who achieve performance better than this threshold are invited to engage with Anthropic's recruitment, potentially opening doors for new talent in the field. Overall, this move not only promotes competitive spirit but also aims to enhance the development of AI efficiency and efficacy.