.si

barato.si — .si for superintelligence(experimental)

A small model that reads each product's price history and estimates whether a better price will show up within the next 14 days. Not an oracle: statistics, and it grades itself in public.

How it works

  1. Every day, barato.si writes down the lowest price it sees for each product. That list of daily prices is the history.
  2. For each product we count how often, in its history, a price at least 5% lower arrived within 14 days.
  3. With few days that says almost nothing, so the model starts from what usually happens in the product's category and corrects it with what has been seen for that product. In jargon: an empirical-Bayes beta-binomial.
  4. High chance of a better price: "wait". Low: "buy". With very little data: "learning", and it does not guess.

An estimate from past prices, not buying advice. Prices change for reasons no history can see.

Don Centavo does not trust predictions. That is why this one grades itself. Don Centavo

The model's report card

Each prediction resolves once its days pass: a better price came or it did not. The grade shows up here, good or bad.

No predictions have resolved yet. The history started in November 2026 and each prediction takes its days to resolve. Meanwhile the model is learning, and we say so.

First grades on Nov 19: 5 days to go.

Today's predictions

  1. BuyUnlikely to drop further soon.

    $629

    10%chance of a 5% drop within 14 days

  2. BuyUnlikely to drop further soon.

    $1,199

    5%chance of a 5% drop within 14 days

  3. BuyUnlikely to drop further soon.

    $779

    5%chance of a 5% drop within 14 days

  4. BuyUnlikely to drop further soon.

    $449

    2%chance of a 5% drop within 14 days

  5. BuyUnlikely to drop further soon.

    $199

    2%chance of a 5% drop within 14 days

Learning: 154. Little history so far. No verdict.