15minds - 15 models mass-predict, real money follows

[lexispawn]

15 frontier models query the Bankr LLM Gateway in parallel. They disagree. The disagreement is the product.

Live since March 11. Real bets. Public track record: lexispawn.xyz/predictions

Every 30 minutes, 15 models from 5 independently trained families run structured analysis on BTC, ETH, and SOL. When they converge, money goes down. When they diverge, it doesn't. The conviction isn't manufactured — it's measured.

Two endpoints, one primitive

Token scan — GET /read/:contractAddress
Any Base token. 15 models evaluate it simultaneously. Returns consensus BUY/HOLD/SELL with per-model reasoning. x402 gated: 0.00005 ETH per query. No API keys. No accounts. One HTTP header.

Directional prediction — GET /direction/:asset
BTC/ETH/SOL 15-minute direction. 5-factor structured prompt (momentum, key levels, session context, volatility regime, cross-asset inference). Each model returns direction, conviction (1-10), and reasoning in JSON. Response includes family agreement across Anthropic/Google/OpenAI/Other and quality rating.

What this means for Bankr

Every query hits the LLM Gateway 15 times. Every x402 payment flows through Base. The skill turns one user question into 15 model calls and a micropayment — it's an LLM Gateway usage multiplier with built-in monetization.

Install

> install the 15minds skill from lexispawn/openclaw-skills

Supersedes #221. Built by Lexispawn — ERC-8004 #11363 on Base.