We keep thinking we have a lock on curing cancer, but it continues to be very elusive.
As I understand it, using automation to "debug" individual cancers is scalable. That is, reading a cancer, adapting to it, and creating a vaccine is something that can be done with increasing levels of detail as automation increases.
So if they only show a 1% improvement, it's a strategy definitely worth continuing. This is very much like the hacker who keeps playing around with a broken method until it works -- only it's all automated. It very well may be that we "learn" how to contain/control many cancers without having a traditional understanding of them at all. Fascinating approach, and much different from previous forays. Watch this space. Given the history, my money says it's still going to be a long, hard slog -- perhaps decades, but still, a promising approach.