Notion page: 366f0841-43c3-80e8-b93a-faaff7c52d0f Last edited: 2026-06-10T12:29:00.000Z
D’après ces données [D1…n] et ces outils [O1…n], ces précédents résultats [X1…n], je peux produire [P1] ces résultats [Xn+1…n+m]
production de donnée D
production d’outils O
production de preuves P
avec X est en fait pareil que D: D = O(D) et P est une preuve / theoreme ou un axiome
production de résultats E qui sont en fait des données D
papier W = set de D,O,P
blockchain… identity tracking (who did what), all versionned, available, can be built upon, it is all references,
To launch this kind of science Project : we could create a database, a tool, a community, a website where people could create DOIs associated with these three or four main categories we highlighted and defined these contributions this way. An agentic system could parse through previously published papers and define, as best as possible, the different D, O, and P elements. And link them together into other citations and other tools.
This would be the dream of translating all mathematics into the formal language lean, but for science, translating all of science into a formal language, and knowledge graph.
easy for automated tools to be applied on
directed graph
groupped: e.g. a tool can be a set of tools composed
a dataset can be a dataset composed or a new version
and then obviously, things can be wrong, data can be fake or wrong, a tool can have a bug, a proof can be wrong. How do we label that, and how do we treat the dependent works? How do we judge the gravity and strength of a relationship? I think we consider it and all its dependents flawed and wrong until re-done to show the link was not required, or an updated version or a different dataset/proof, allows us to have the same result.
How do we judge of research output? it is the number of things created and number of things created on top of yours by others. same as now, almost
how do we define contribution? time spent, for sure. if a PI claims to have spent 50% of their time, we will count how many they did.