A Nature study reported AI-designed protein redesigns that shift the “rules of evolution” in phage-assisted continuous evolution (PACE). The work applies controlled selection pressures to evolve a botulinum neurotoxin protease (BoNT/E) and tests how different AI redesign strategies alter which mutations emerge. The research frames AI as a way to supply a “boosted starting point” so experimental evolution more efficiently converges on higher-function enzyme variants. By comparing redesign approaches under comparable selection pressure, the study focuses on mechanistic differences in evolutionary trajectories. — Key takeaway: the results reinforce AI-guided starting conditions as a practical lever for accelerating functional protein discovery, especially when combining directed evolution with controlled selection systems.