FreqAI is the prediction box. The layers around it are what we actually built.
CORE
FreqAI × LightGBM, 160 pairs
5-minute candles on MEXC spot, models retrained every 12 hours per pair,
2-hour-ahead return regression. Pairs are vetted by a data-completeness pipeline
(listing age, volume, candle history) before they ever reach the whitelist.
CORE
Dual-LLM advisory gate
A local gemma model tags the market regime every hour; Claude writes a risk
directive three times a day. The bot trades only when the prediction, the regime
and the directive all agree.
NEW
AI trade post-mortems
Every closed trade is autopsied by an LLM against measured facts (max adverse
excursion, pre-entry pump, exit reason). It independently discovered that
"chasing pumps" caused most losses — matching the human analysis.
NEW
Nightly LLM researcher
Every night Claude reads the bot's own performance dossier and proposes testable
hypotheses in a constrained DSL. Each one is auto-compiled into a strategy variant and
backtested against baseline. Only proven winners are surfaced for human approval.
NEW
Semantic news quarantine
An LLM reads crypto news every hour. Hack, exploit or delisting headlines
auto-quarantine the affected pair for 24h — the kind of signal a numeric model
can't see by design. Positive signals run in shadow mode until proven.
DISCIPLINE
Backtest kills bad ideas
On its first night the researcher proposed a plausible threshold change that
backtesting showed would have lost 7,000+ USDT (paper). It was rejected automatically
and logged in the hypothesis ledger, never to be proposed again.
