This course interprets the public Boost guidance at revision 9de51d8, a public set of instructions an assistant can follow when it investigates a problem. The examples, diagrams and exercises are original. No private task packets are included.
FunSearch uses model-generated programs, automated evaluation and a diverse pool of candidates. AlphaEvolve uses evaluators and a program database to guide program evolution. AlphaDev is a trained reinforcement-learning system that searches assembly instructions. These systems are research results; they do not show that the method in this course improves your work.
The exercises use observations written for this course. No agent experiment runs automatically. Your JavaScript runs locally in a disposable worker. The fresh cases are unshown in the interface until submission, but they're in the download. Passing them doesn't certify research judgment or learning.