Viral infectious diseases remain a major threat to global public health, highlighting the urgent need for broad-spectrum antiviral agents and rapid-response drug discovery strategies. Artificial intelligence (AI) and other emerging technologies are accelerating antiviral drug discovery. Based on SARS-CoV-2 M^pro structure, we designed DC402234, a potent broad-spectrum antiviral candidate that does not require co-administration with ritonavir and possesses robust activity against multiple clinically variant strains. DC402234 is currently undergoing international multicenter Phase II/III clinical trials. To further expand broad-spectrum antiviral discovery, we performed computational analyses of 12 groups of 3C/3CL proteases representing the entire Pisoniviricetes’s diversity. We then developed a deep learning-guided bidirectional molecular evolution strategy and discovered DC410260, the first pan-Pisoniviricetes inhibitor targeting 3C/3CL proteases. DC410260 showed potent inhibitory activity against proteases from all 12 groups, broad antiviral efficacy against 9 distinct viruses, and significant in vivo antiviral activity. In addition, we solved 7 co-crystal structures of DC410260 in complex with representative 3C/3CL proteases, thereby elucidating its molecular basis of target recognition and inhibition. Collectively, these studies not only provide promising broad-spectrum antiviral candidates, but also establish a new paradigm for AI-enabled antiviral drug discovery.