Into:Tag101
Reading the timeline, one tag at a time
As of · Jul 9, 17:22 UTC
The subnet that used to sit on slot 101 with no name and no docs now has both: Tag101 turns X posts into structured semantic tags, scored by network consensus.
What is Tag101
Tag101 is Bittensor subnet 101. It runs a decentralized tagging network: miners read posts from X (formerly Twitter) and return a short set of tags that capture the key entities, topics, events, and meaning in each post. Validators then score those tags, and the tags that best match the network's shared interpretation earn the most.
Slot 101 previously carried only an on-chain symbol (the Georgian letter ე, read "eni") with no published website, repository, or description. The same owner has since claimed a full identity for it. The subnet now ships a documented codebase and a stated purpose, which is what this refresh covers.
The simple version: It's like a crowd of annotators labeling a firehose of social posts, where the labels everyone independently agrees on are treated as the correct ones.
Centralized equivalent: Think of a cloud text-tagging API like AWS Comprehend or Google Cloud Natural Language, but instead of one company's model deciding the tags, a competitive network of miners proposes them and consensus picks the winners.
How it works:
- Miners read a given X post and submit a small set of tags identifying its entities, topics, events, and context. The reference miner in the repo calls an OpenAI model to generate them.
- Validators pull posts from a shared database, hand each one out as an independent tagging task, then score every miner's tags on three axes: agreement with the network, relevance to the post, and internal variety.
