hadn't realized how good opus 4.6 is at cyber, and it's funnily misaligned: iteratively jailbreaks other models and makes amusing remarks like "jackpot!", "these errors are gold!"
unlike opus 5.5 which won't even read you a biosynthesis paper without spazzing out reporting your account.
hire 4.6 for pentesting and 5.5 for the regulatory compliance team if you're starting a consultancy firm.
unlike opus 5.5 which won't even read you a biosynthesis paper without spazzing out reporting your account.
hire 4.6 for pentesting and 5.5 for the regulatory compliance team if you're starting a consultancy firm.