Over the last two weeks, Love-Scout got a meaningful upgrade focused on one goal: helping more Hive creators feel seen, welcomed, and supported.
This update includes better language support, stronger reliability checks, and a new retention-focused follow workflow.
Better support for non-English intros
Love-Scout can now detect more languages and choose the best match more reliably when a post contains mixed signals.
That means new authors are more likely to get community pings and recommendations that actually match how they write.
In simple terms: fewer wrong guesses, better first impressions.
Smarter recommendation coverage
We expanded recommendation data so replies can better match different interests, including newer finance-focused pathways.
The goal is to reduce the "now what?" feeling for new users by giving them more relevant people and communities to explore.
New: Follow At-Risk creators (retention workflow)
This is the biggest thing I have not announced yet.
Love-Scout now has a dedicated script that identifies creators who may be drifting away and follows them in a measured way.
What "at risk" means
The list is generated from Hive on-chain history using HafSQL and compares two time windows:
- Previous 30 days (their baseline activity)
- Last 30 days (their current activity)
Creators are flagged when their recent posting rate drops sharply versus their own baseline.
How the query works (plain English)
The workflow looks for creators who:
- were previously active enough to matter (not one-off accounts)
- posted significantly less in the last month than the month before
- still posted somewhat recently (so they are still reachable)
- are unlikely to be high-volume automated accounts
This gives a practical "high-risk but recoverable" segment.
Why follows?
A follow is lightweight, but it creates a visible social signal:
- "Someone noticed your work."
- "You are not shouting into the void."
For creators on the edge of disengaging, that tiny signal can be enough to bring them back into normal activity.
Safety and pacing
The script follows in small batches with short delays, so actions stay controlled and respectful of chain limits.
Reliability improvements behind the scenes
We also tightened internal behavior to reduce accidental regressions:
- Language detection now scores all candidate languages before picking one
- We added focused regression tests for language detection and follow operations
- Integration docs were cleaned up so runbooks match real code paths
Why this matters
Love-Scout is moving from "welcome bot" toward "welcome + retention assistant."
Welcoming newcomers is important.
Helping people stay is even more important.
These updates are a step in that direction.
Feedback?
What do you think? How can we make this better?