Awesome project! Wanted to let you know that Bombshell started running this as a GitHub Action in CI.
When I ran a backfill against open PRs in the Clack repo, the results seemed pretty accurate! We had one false positive, one real positive, and the rest were correctly detected as organic.
The false positive was:
Classification: automation (score: 45)
| Signal |
Points |
Detail |
| Frequent repository creation |
+25 |
8 repositories created in a short timeframe (within 24 hours) |
| 24/7 activity pattern |
+30 |
Active 20/24 hours, 2h max rest, 10.0 events/hour |
It'd be interesting to see if you could eliminate this with some fine-tuning of the weights here? Repo creation and 24/7 activity seem like fairly weak signals that would flag many people who work in open source.
Awesome project! Wanted to let you know that Bombshell started running this as a GitHub Action in CI.
When I ran a backfill against open PRs in the Clack repo, the results seemed pretty accurate! We had one false positive, one real positive, and the rest were correctly detected as organic.
The false positive was:
It'd be interesting to see if you could eliminate this with some fine-tuning of the weights here? Repo creation and 24/7 activity seem like fairly weak signals that would flag many people who work in open source.