Recaster

The Recaster Miniapp is ready. Currently, it offers three features:

- Search and view user stats, including Warpcast label, Neynar score, and OpenRank.
- Compose and schedule casts (notifications are coming soon).
- Purchase Warpcast invites directly with crypto.

This is a fun experiment, use Recaster within Warpcast, will add more features if people like it.

The Recaster Miniapp is ready. Currently, it offers three features:

- Search and view user stats, including Warpcast label, Neynar score, and OpenRank.
- Compose and schedule casts (notifications are coming soon).
- Purchase Warpcast invites directly with crypto.

This is a fun experiment, use Recaster within Warpcast, will add more features if people like it.

https://app.recaster.org

My followers number looks off, it’s less than Warpcast. You doing some type of filtering?

Building /recaster and other farcaster tools

a frequent problem we heard was other clients showing much higher follow and reaction counts compared to warpcast

turning on the `x-neynar-experimental` header will now filter out lower noise from these counts, the header is _not_ turned on by default so developers must explicitly choose to use it

- this is experimental by definition, we will continue to improve the filtering behind the scenes
- the numbers are not expected to match warpcast, they are meant to be an overall noise filter

to read what all the experimental flag does, read here: https://www.notion.so/neynar/Experimental-Features-1d2655195a8b80eb98b4d4ae7b76ae4a?pvs=4 and reach out with any feedback 🪐

https://warpcast.com/rish/0x6ea87140

hmm. will change back

fwiw warpcast filtering isn't 100% accurate so if your goal is to filter noise, it misses a bunch of that 

so depends on whether the developer goal is to
- show protocol data
- reduce noise
- match warpcast (not doable)

building farcaster infra @ /neynar 🪐 | casting @ /rish

we have a work item to start putting the user score in the user object without requiring the header. should happen in the next week or two. then the header will be for the filter only.

How can I get neynar score without setting the experiment header?

Usecase: show people full follower count and neynar score at the same time.

its different for different things e.g. follower count vs reply count and we have to tweak the threshold if the underlying model updates significantly

Have you guys done any internal testing? I plan on doing my own but I’m curious if Neynar has a specific score that you know does a very good job at filtering noise