Why isn't all human knowledge in one big JSON file? Guidelines are decision trees, but they're not written to be applied by rote, because the identical patients with identical cancers posited in the hypothetical don't exist. The guidelines are not written for maverick clinician movie protagonists to navigate the decision tree in real time while racing against an oncological clock, they're for teams of clinicians who are very well trained and used to working with each other, and who have the skills to notice where the guidelines might be wrong or need expansion or modification. That is, they're abstractions of the state-of-the-art.

Now it'd be nice if these could be treated like source code on some sort of medical VCS, to be easily modded or even forked by sub-specialists. But wetware dependencies are way more complicated than silicon ones, and will remain harder to discretize for some time to come.

It's not that the author's aspirations misguided, they're great. But I believe progress in this area is most easily realized as part of a relevant team, because what looks conceptually simple from outside the system only seems that way because of a lack of resolution.