Repository navigation
Likes and Ranking
A star tells you someone was interested once. A Like on DevStream is somebody saying "I use this", and the list is ordered by them. What makes that order worth trusting is simple: one account, one Like, and the accounts have to be real.
Every listed repo has its own thread in the Likes category of Discussions. Select Like on the leaderboard to open it, then react to the post with 👍.
That's the whole thing. To take a Like back, remove your 👍.
| Counts | Doesn't count |
|---|---|
| A 👍 on the entry's post | Any other reaction, such as ❤️ or 🚀 |
| One per account | A second 👍 from the same account |
| From a trusted account, owners included | Reactions on comments in the thread |
| On the thread the bot opened for that entry | A thread someone else opened about the repo |
Every Like that counts is worth exactly one. Nobody's Like is weighted up or down, and that includes the repo's owner and me.
Only the bot can open threads in the Likes category, and only its threads are read. Anyone can comment on an entry's thread, so it's a good place to say what you use the repo for.
The bot counts every entry's 👍 reactions once an hour, then rebuilds the site. A new Like shows up within about an hour.
The Top tab, and each category list, is ordered by:
- Likes, most first.
- GitHub stars, when two entries have the same number of Likes.
- Name, A to Z, when those tie too.
Next to each rank, the arrow shows how far the entry has moved over the past week. The Likes column shows how many it gained. Both come from the daily snapshot the bot saves, the same comparison Trending uses.
Entries with a health warning stay listed and ranked until they move to the archive. Hidden and archived entries aren't ranked.
The Trending tab ranks entries by the Likes they gained over the past week. The bot compares today's count with the snapshot from a week ago, or with the oldest snapshot from the past week if that one is missing. The rank arrows and the Likes column use the same comparison.
A repo needs at least 5 new Likes that week to appear. Newly submitted repos join Trending once they've been listed for a week; repos from the launch seed list don't wait. Flagged repos stay off Trending until I've reviewed them, and so does an entry whose name a different repo has taken over, until the health check sorts it out.
Newest entries first, by the date they were added to the list.
If you've reacted and the count didn't move, it's almost always one of these:
- It hasn't been an hour yet. The count updates hourly.
- Your account is under 30 days old. Likes from new accounts don't count. Once the account is 30 days old, your Like starts counting on its own, though it can take up to a week. You don't need to react again.
- Your account has no public repos and no contributions in the last year. Private contributions only count if your profile shows them. Trust Checks explains how to turn that on.
- You used a different reaction, or reacted to a comment instead of the post.
- You're on a bot or blocked account.
Your 👍 stays on the thread either way. If your account becomes trusted later, it counts from then on.
Sock puppets are the obvious way to cheat a list like this: make a pile of accounts, Like your own repo with all of them. Three things get in the way.
The trust rule. Accounts younger than 30 days, and accounts with no activity at all, don't count. A fresh batch of throwaway accounts moves nothing.
Burst detection. A sudden spike of Likes, mostly from accounts that pass the trust rule but are only a few months old, is what bought votes look like. When the bot sees one, it flags the entry and opens an issue for me to review. While the flag is on:
- the entry stays listed and ranked as normal;
- its Likes still count;
- it's held out of Trending.
The flag stays on until the spike has passed and I've closed the review. Only Likes that count can set it off, so nobody can get someone else's entry flagged by aiming throwaway accounts at it.
Drip detection. Spread the same pattern over a longer stretch to stay under the burst check, and the bot spots that too. A drip doesn't flag the entry. It opens a review issue, and I take a look.
When a review turns up sock puppets, those accounts go on a private exclusion list, and their Likes stop counting on every entry. Accounts that are part of the abuse can also be blocked from the organization.
I don't publish the exact numbers behind the burst and drip checks, since that would only tell people how to slip under them. Found a way past them anyway? That's a security issue, and I'd like to hear about it privately. See Security.
- Trust Checks: the full trust rule.
- Removal Policy: what happens to Likes when an entry moves to the archive.
- FAQ: more questions about Likes.
Site · Repo · Discussions · CC0 license
Getting listed
Likes
Keeping it clean
Help
The project