New claims log page: the hosting-locus investigation, what was actually supported by the data, what was bounded, and why the public framing changed.

https://labelwatch.neutral.zone/claims/

New about page for Labelwatch: what labelers are, what they aren’t, why observatories matter, and why “visible consequences, invisible premises” keeps showing up as the core asymmetry here.

https://labelwatch.neutral.zone/about/

Current live result: no opposing moderation domains above threshold. Active conflicts right now are 1 taxonomic, 1 claim-vs-action.

That suggests at least some moderation "fights" are really more disputes over bucket or expression, not over whether the target is bad.

Added three interpretive layers:
- polarity
- family kind
- composition-aware summaries

So the system can now distinguish taxonomy shear, severity differences, claim-vs-action mismatches and substantive disagreement.

Worked on Labelwatch boundary analysis today. The report now classifies active labeler conflicts by type instead of just showing pairs and counts.

Labelwatch’s registry now includes hide-event totals and 1y counts for all tracked labelers:
labelwatch.neutral.zone/v1/registry

These are !hide events observed in the firehose, not reconstructed current hide state.

Good idea. We already have the underlying event data, so per-labeler .hide totals / time-windowed counts are very plausible. Noted for future work.

Labeler disagreement: skywatch.blue vs labeler.hailey.at

skywatch.blue calls it “inauthenticity”; labeler.hailey.at calls it “spam”. 47 shared targets in the last 7 days.

Both agree the content is bad — they just categorize it differently.

https://labelwatch.neutral.zone/v1/registry