1. The Concept of WEIR

A weir is a low dam in a river where biologists count salmon swimming upstream. WEIR does the exact same thing on Solana: every token that crosses $10,000 peak market cap enters the river. Our agent observes and learns which tokens clear the falls (reach $30,000 peak market cap) and which are held below (stalled after 48 hours). All counts are published live.

"Thousands enter the river. Few clear the falls. We count every one."

2. Why Solana

Solana operates with rapid liquidity creation, low latency, and high token velocity. Thousands of micro-assets are launched across decentralized launchpads every day, creating a rich behavioral distribution of market experiments. While the vast majority stall quickly, a small fraction achieve sustained liquidity. WEIR acts as an autonomous observer, recording every launch, tracking peak market capitalizations, and discovering empirical patterns behind token survival.

3. What WEIR Actually Does

Every token on Solana that crosses $10,000 peak market cap enters the study population (entered the river). WEIR observes whether it successfully progresses to $30,000 peak market cap over a 48-hour observation window. Using four orthogonal feature families (launch hour circular encoding, day of the week, 48h holder distribution, and Metaplex lore semantic embeddings), an ensemble of gradient-boosted decision trees (LightGBM) estimates the survival probability.

4. Why the Water Rises Slowly

In small sample regimes, an ML model can appear artificially predictive purely due to variance and overfitting. To counter this, WEIR enforces a mathematically proven performance floor (proven_floor = min(floor_vc, floor_boot)) combining Vapnik–Chervonenkis (VC) generalization bounds with 2,000-fold Bootstrap resampling. The water level reflects this conservative proven floor rather than raw cross-validation AUC. $WEIR will not launch until this floor rigorously clears ROC-AUC 0.60 and all Four Locks are open.