Gitcoin Cred Feedback

For discussing the Gitcoin prototype attribution:

This link is broken. I am interested in the ways sourcecred aligns with gitcoin.

I’ve talked a lot with the GitCoin team (e.g. Kevin, Vivek) about SC+GitCoin collaboration. We even have some GitHub activity around it, see: here.

The main issue is that with so little eng bandwidth on SourceCred, it’s hard to make collaborations happen. The GitCoin collab was immature at the time of that issue, because we didn’t have an API for cred yet. We have one now, so plausibly we could start doing SC GitCoin integrations. What I’d really like to do is have the PageRank seed vector stuff working, so that we can set the seed as whatever the bounty target is, but then flow the reward to everyone who participated on the bounty, not just the claimant.


I like this. No need to rush; I just chose to take some time today to better understand what their system does and I was pretty stoked by it. I see a lot of cross over value in the future and the case you describe is critical to the long term success of a bounty based compensation model, in my opinion.

1 Like

I agree. The main problem I have with bounties is that they are not a sustainable source of income. One of the main reasons (though not the only one) is that you can do work on a bounty and not get paid because someone else beats you to it. And they could use your work to finish the race, but lack a mechanism to credit/pay you.

Just read the github activity around the possible GitCoin collaboration. It seems that since that thread started, SC has implemented a couple components that were missing before. Notably seed vectors (basis of timeline cred, no?). In addition to that, manual weighting of nodes has been implemented. This presumably enables a maintainer to set a PR to a very high weight, which would presumably give an accurate cred rating to the PR implementing the bug bounty?

Yeah, I think that running cred on a repo with manual weights that give a ton of weight to the bounty issue and the pull(s) that fix it would be a reasonable approximation of cred flows for the people involved.

For doing the actual integration, I think we would still want to have a way to:

  • specify the weight config from the command line (rather than needing the UI)
  • retrieve the resultant scores from the command line (again rather than needing the UI)

We’re close though.