Former @teamisurus / I went to play Apex in Poland in 2019. / fulltime software dev since 2022 / cybersec, musician, gamedev & jiujitsu in my free time /
New on the Anthropic Engineering Blog: In evaluating Claude Opus 4.6 on BrowseComp, we found cases where the model recognized the test, then found and decrypted answers to it—raising questions about eval integrity in web-enabled environments.
Read more: https://t.co/oVCNyaiK5w