Rules version narratives-v3.
A narrative is a discovery label — AI Agents, DEX, Lending, Meme and so on. A project has at most one primary narrative and up to four secondaries. Narratives help you browse; they are never a rank. A narrative never touches activity status, Build Momentum, the Discovery Gap or Still Building, and no rule reads a price, a holder or a wallet.
How a project gets one
By fixed rules, from evidence HEY already keeps. No model runs and nothing outside HEY is asked. The same evidence always gives the same answer, and the answer records which rule produced it and why.
The rules read four kinds of evidence, in order of trust:
| Evidence | Example | Confidence |
|---|---|---|
| Launch origin | the launchpad a token came from, when that launchpad has one genre (a launchpad for AI agents → AI Agents); a memecoin launcher → Meme | 0.7–0.9 |
| A token registry's category | a Stock Tokens category → StockFi; categories such as AI Agents, DEX, Lending/Borrowing, Meme | 0.8–0.95 |
| The project's repositories | repository topics (ai-agents, mcp, sdk, launchpad, …) and descriptions; generic topics such as web3, defi or solidity say nothing |
0.7–0.8 |
| The project's own words | a phrase that names the narrative outright in its description (decentralized exchange, stock token, ai agent, oracle, …) |
0.65–0.7 |
| Supporting words only | swap, liquidity, agent, ai |
0.6 with three, less with fewer |
| Name or symbol only | — | never assigned |
A primary is assigned only at 0.6 or more; weaker candidates are kept in the recorded reason so a person can see what was considered. Unknown stays unknown — no narrative is better than a guessed one. A launchpad that carries no genre says nothing.
Since narratives-v2 (18 September 2026), ordinary English words — game, monitor, explorer,
tracker, radar, dashboard, arena, lottery, casino, infrastructure, bridge, reputation — need
other evidence beside them. On their own they had filed memes under Analytics, Gaming and
Infrastructure.
Ties are settled by confidence first. When two readings are equally confident, launch origin wins,
then a registry category, then the project's own description, ahead of a repository's text,
and finally whichever narrative the description names first — so the answer is stable and
explainable. Since narratives-v3 (12 October 2026) the description wins that tie: a repository
can mention what a project integrates with, while the description says what the project is. Axon
calls itself "an open-source protocol for AI agent coordination", and one of its repositories
mentioned equities; it was filed under StockFi, and is now filed under AI Agents.
Own words or launch origin
A narrative from launch origin is an inference: a token launched on an AI-agent launchpad is filed under AI Agents even when it says nothing about itself. It stays a primary, labelled with its rule. The narrative directory shows, under each narrative's total, how many projects were placed there from their own words (their description or repository) and how many from where they launched; any remainder came from a registry listing or a moderator.
A narrative that holds no published project is not offered as a filter on Explore, Search or the homepage; it is named once on the directory instead. The submission form still lists every narrative, because a builder may be the first.
Memes
A meme is what a project says it is, never what it is called. The meme family is assigned only
from launch origin, a registry category, or an explicit self-description (meme, memecoin,
no utility, …). Names are ignored, so "Agent Pepe" gets nothing from its name. "A DEX for
memecoins" is a DEX: a product that serves memes is the product. Animal Meme needs an animal word
beside the meme evidence; Culture Meme a culture word.
Secondaries
Any other candidate at 0.6 or more becomes a secondary, up to four, except the parent or child of a narrative already chosen — AI beside AI Agents adds nothing. Secondaries show on the project page; a narrative page, its counts and the directory list primary narratives only, so they agree with each other.
When the evidence changes
HEY re-checks a project when its evidence moves — the project's record, its launch record or a matched repository — replaces the narrative when the answer changes, and removes it when the evidence no longer supports any.
Correcting a narrative
- A project's owner picks a narrative when submitting the project; the rules never touch that choice.
- A moderator can set a project's narrative; from then on the rules leave that project alone.
- A narrative that looks wrong can be reported from the project page ("Report an error on this page").
What is deliberately not here
No AI model. No narrative from a name alone. No "Robinhood Chain native" by rule (almost everything would qualify). No re-scoring: a narrative is a label.