How Phasix works under the hood
Every scan hits three data layers simultaneously.
Layer 1 - Live holder topology
We query the token ledger directly and compute real-time holder concentration, top wallets share as a percentage of circulating supply. Not cached. Not from a database. Current ledger state at query time.
Layer 2 - Market microstructure
We pull the highest-liquidity trading pair and extract: volume velocity (1h volume vs. 24h hourly baseline), buy/sell transaction ratio across multiple time windows, price acceleration per timeframe, and liquidity depth. These signals are composited, not displayed raw.
Layer 3 - Phase classification
A deterministic signal tree runs against all fetched data. Holder concentration thresholds, velocity multipliers, buy pressure ratios, and pair age gate each classification. No black box, pure conditional logic on real on-chain state. Output is a lifecycle phase and a Window Score (0β99) derived from liquidity-weighted buy pressure.
Layer 4 β AI synthesis
The full signal set is passed to a well-trained language model running inference on the edge. It produces a focused read on what the on-chain state means for that specific token in that moment. You can ask follow-up questions, the model has full context of the scan.
No predictions. Just the current market structure of a token, read from source and verified by backtested data.
https://t.co/sI12z9CEZ6