Crypto-LLM
LLM decisions for crypto markets, with hard risk limits.
Making a model’s judgment checkable
Crypto-LLM is an experiment that combines a large language model with crypto market data. It brings together prices, technical indicators and account state, produces an analysis and a decision record, and tracks what happened next.
What the system includes
- Market data and more than 20 technical indicators, given to the model as structured input.
- Hard limits on position size, leverage and drawdown, with a log of every trade and decision.
- A test environment and historical records for reviewing decisions, so the gap between the model’s analysis and actual results stays visible.
The questions behind it
Why did the model reach this judgment? Do the conditions it relied on still hold? When something goes wrong, how does the system limit losses and stop? These questions shaped the risk controls and the review process.
This page documents the tool’s design, development and experiments.