Methodology: code first, the model explains

The idea

A language model writes fluent text, and fluent text can contain a wrong number. Cryptokly separates the two jobs. Deterministic code computes every number; the model receives those numbers and explains them. The explanation stays attached to figures that the code produced.

What the code computes

For each pair the engine reads candles on five timeframes (1M, 1W, 1D, 4H and 1H for spot; 1D, 4H, 1H, 15m and 5m for futures and tokenized stocks) and computes, from closed candles, the swing highs and lows, the trend word, and the support and resistance zones. It adds the Wilder RSI(14), which includes the forming candle, and the ATR(14) behind the zone width and the volatility word.

Because the same code runs for every pair and every request, the same candles produce the same numbers each time.

What the model does

The model you pick receives the computed figures and writes the explanation: what the trend words say together, where the nearest zones sit, and how the RSI reads next to the trend. It describes the chart as it stands and its text is limited to the supplied figures.

How to read an answer

Treat the card as the source of numbers and the text as the explanation of those numbers. When the text and the card ever differ, the card holds the computed figures. A timeframe with too little history shows "History too short", and no reading is invented for it.

Where the data comes from

Prices and candles come from the spot markets of major crypto exchanges. Public pages such as the coin pages and the lessons use the same hourly engine run, and each live example shows the UTC time of that run.

What this offers

Cryptokly describes charts and teaches how to read them. It offers education and description in place of advice, and crypto trading carries a real risk of losing money, up to the whole amount invested.