Early discovery
Source-first detection of brand-new tokens, the discovery sources per layer, and the early-token processing pipeline.
The system must not depend on CoinGecko, CoinMarketCap, curated lists, delayed aggregators or social posts — they detect too late. Discovery is source-first: first-party blockchain and exchange events.
Required event categories
- contract deployment
- token-mint creation
- launchpad token creation
- bonding-curve creation
- first liquidity addition
- DEX pool creation
- first swap
- migration init / complete
- CEX listing announcement
- CEX trading activation
- abnormal transaction growth
- abnormal liquidity growth
Discovery sources
| Layer | Monitored |
|---|---|
| Blockchain | Per-chain adapters (Solana, Ethereum, Base, BNB Chain, Arbitrum) — events differ per chain |
| DEX | Pool creation, reserve changes, swaps, liquidity adds/removes, pair creation, migration |
| Launchpad | Token creation, bonding-curve progress, buyer growth, migration state, creator activity, completion/failure |
| CEX | Listing announcements, schedules, deposit activation, market activation, order books, trades, suspensions |
| Social | Supporting evidence only (X, Telegram, Discord, Reddit, YouTube, official channels) — never the sole trigger |
The processing pipeline
- New chain / exchange eventDeployment, pool, swap, migration, listing
- Event normalizationCommon internal schema
- Token identity resolutionContracts ↔ symbols ↔ profiles
- Lifecycle updateState change → immutable event
- Security screeningPermissions, creator, rug mechanisms
- Tradability + sellabilityCan it be bought — and sold?
- Liquidity + wallet/holder analysisReserves, slippage, clusters
- Early momentum + social enrichmentAcceleration + narrative
- Memory retrieval + strategy matchCompare to past launches
- Scoring → decisionIgnore / watch / simulate / review / execute
Runs on incomplete data
The full process must complete even when little is known. A two-minute-old token is evaluated with whatever exists — uncertainty is quantified, not assumed away. See data maturity.
Each chain needs its own adapter because creation, liquidity and transaction events differ. Collectors stay separate so one failing source never stops the discovery pipeline.