Retrospect vs Prospective Cred

From what I gather, currently the app is aggregating historical data from the repo graphs to asses how a finite pool of cred could be distributed across all the actions and contributors at the current state.

So this lead me to wonder, this feels a bit historical and retrospective – have you considered or thought about ways to it could be more prospective or predictive? Two scenarios which came to mind for me were:

  1. If you made a new contribution, could you predict or estimate how much cred it would accrue? [like which areas of the repo would be valuable to concentrate effort on?]
  2. If a new pull request was logged, based on say characteristics or location in the graph, could SourceCred suggest which collaborators might be best suited to address it? [i.e. obvious one would be for a bug whoever originally worked on it, but I can imagine many much more nuanced answers to this…]

While we haven’t yet discussed predicting how much cred a contribution will make at the algorithmic level (afaik), the main mechanism for accruing value to the Grain token is Boosting, which creates a prediction market on which contributions will increase in cred over time. So players of the “SourceCred game” will be financially incentivized to use whatever means at their disposal (including algorithms presumably) to make predictions, creating “prospective” cred.

I haven’t seen this talked about yet (doesn’t mean it hasn’t:) But the graph is a rich source of data for all sorts of analysis. Plenty of ideas generally around that.


super interesting, I think my efforts to merry this with Augmented bonding Curves requires this prospective (e.g. defined bounties that require fulfillment) ~ I have a feeling that boosting is or could be extended to be used like token curation, even creating bounties and boosting them - once they are fulfilled, are in use, and create usage value, that revenue goes through reserve to mint tokens (grains or named differently in different instances), which then get redistributed to cred holders, and via this adjust weights - like backpropagation :slight_smile: @s_ben please do join our call on June 11 7PM CEST and I’ll join the next community call, would love to grasp this better. I think with the TE book project we have a relatively simple experiment to test these variations out , linking here again: