EchoLikeEN

← Volver a Directorio

Actualizado 23 oct, 9:23 a. m. Bogotá

Cómo se construyen estos datos

Las reglas que sigue cada tabla de este directorio. Se revisan automáticamente en cada construcción y con un revisor de IA independiente antes de publicar. Cuando algo está mal, la corrección se convierte en una regla nueva aquí.

Los nombres de columna y los valores se publican en inglés: el directorio mantiene un solo estándar de datos, así un valor significa lo mismo en todas las versiones de idioma.

  1. Each column holds one fact: one value, or one list answering "which X" from a fixed vocabulary (for example which chat apps, which memory stores). No umbrella labels that mix several different things.
    Por qué: A mixed column can't be filtered, compared or checked; a fixed-vocabulary list still answers one question (which X).
    Añadida 2026-10-22, después de: Coco: fixed-vocabulary lists are one fact
  2. Every value links to a source page that states it, or is calculated from other sourced values with the formula shown in words.
    Por qué: Readers must be able to check every cell themselves.
    Añadida 2026-10-22, después de: Nova: hand-verify values with source URLs
  3. Never fill a cell with an assumption or estimate. Unknown stays empty and shows as a dash.
    Por qué: One invented number makes the whole table untrustworthy.
    Añadida 2026-10-22, después de: Nova caught an invented 'typical monthly cost' of $60 for 7 harnesses
  4. Prices come only from the official pricing page, with the exact dollar amount, billing period, date checked and link.
    Por qué: Prices change and third-party summaries are often wrong.
    Añadida 2026-10-22, después de: Nova caught invented pricing data
  5. Every number shows its unit, such as $/month, MB, hours or %.
    Por qué: A bare number like 20 means nothing without a unit.
    Añadida 2026-10-22, después de: Nova: typed units in the data standard
  6. Columns with choices use a fixed list of values, each with a plain-English name. Never "varies" or "depends".
    Por qué: Vague words can't be filtered and hide missing research.
    Añadida 2026-10-22, después de: Nova: closed vocabularies in the data standard
  7. A value checked by a person is never overwritten by AI. If AI disagrees, the disagreement goes to a review queue.
    Por qué: Automatic re-extraction had overwritten a hand-checked value (pi's model providers).
    Añadida 2026-10-22, después de: Nova: mark each value as checked by a person or by AI
  8. Every column has a plain-English name and a one-sentence description, shown in the column list and when you hover or tap its header.
    Por qué: Readers can't use a column they don't understand.
    Añadida 2026-10-22, después de: Nova: the 'Writes its own memory' column had no description
  9. No unexplained symbols. Calculated columns say "(calculated)" and explain the formula in words.
    Por qué: An unexplained mark confuses readers instead of informing them.
    Añadida 2026-10-22, después de: Nova: an unexplained marker on calculated columns was confusing
  10. The page is in plain English. No internal code names (words joined with underscores) in headers, values or text.
    Por qué: Internal names joined with underscores mean nothing to readers.
    Añadida 2026-10-22, después de: Nova: use human labels in headers, not code names
  11. Tables work on a phone (390 px wide). By default about 8 columns show, with column groups and a column chooser.
    Por qué: Most readers open links on their phone.
    Añadida 2026-10-22, después de: Nova: short default column set with mobile-first group toggles
  12. Before calling work done, look at the published pages the way a reader would, on a phone and a desktop, not just at the code.
    Por qué: Code that runs can still produce pages that confuse.
    Añadida 2026-10-22, después de: Nova: our work was careless (a missing description and an unexplained marker)
  13. An independent AI reviewer, using a different model from the one that built the page, reads it as a skeptical first-time reader. High-severity findings block publishing.
    Por qué: Reviews are done by AI models, not by Nova (EchoLike's founder), and by different models from the one that built the page.
    Añadida 2026-10-22, después de: Nova: reviews must be done by models, not by him
  14. Every correction from Nova becomes a rule on this page, checked on every build from then on.
    Por qué: The same mistake should never come back.
    Añadida 2026-10-22, después de: Nova: append every correction to the checklist
Published with deed.page