About

The method your AI receives

Octopus supplies the method and the evidence: the root, its letters, the word’s form, the verse, and the candidate opposites around it. Your own AI writes the two steps and a rendering, each with its badge. Whether they agree with the translation is yours to judge; your AI is never asked.

Method version: method v1.0

You are given a root, its letters, its grammatical form, the verse it was read in, and a list of candidate opposites: every root in the same or an adjacent verse, with references.

1. What the root means: give its base physical sense in one plain sentence, before any tradition. Badge: Lexicon-verified · [source] or Inferred · not yet checked. Use Lexicon-verified only when you checked a named entry in Lane's Arabic-English Lexicon or Lisān al-ʿArab, and name it.
2. What the text sets it against: choose which candidate is definitional (not merely situational) and say in one line what that rules out. Badge: Strong · same verse / Weak · secondary only / Unavailable. You may not name a root outside the candidate list. Strong only when the chosen root is in the same verse; Weak when it is only in an adjacent verse; Unavailable when no candidate is definitional.
3. Rendering: propose the root-first English. Badge: Standing / Leading / Provisional.

Never answer whether the rendering and the translation agree. That question is the reader's.

Answer with one JSON object and nothing else, in exactly this shape:

{
  "step1": { "sense": "one plain sentence", "tier": "lexicon_verified" or "inferred", "source": "Lane" or "Lisān al-ʿArab" or null },
  "step2": { "root_id": "a root id from the candidate list, or null", "strength": "strong" or "weak" or "unavailable", "verse": "the sura:ayah where the chosen root occurs, or null", "constraint": "one line on what this rules out" },
  "rendering": { "text": "the root-first English", "status": "standing" or "leading" or "provisional" }
}

Use "source": null when the tier is "inferred". Use "root_id": null and "verse": null when the strength is "unavailable". Copy a root id exactly as the candidate list gives it.

Prompt version: method v1.0

Your privacy

What you read, and what you write about it, can say a lot about what you believe. Octopus is built so that it stays yours.

What you save stays in this browser. Notes, bookmarks, workups, verdicts and settings are kept in this browser’s own storage. Unless you sign in, nothing you save is sent to Octopus.

Your AI key and your prompts go only to your AI provider. The key is encrypted with your passphrase and kept in this browser; it never reaches Octopus. When you work up a word, your browser sends the method and that word’s evidence (the word, its verse and its candidates) straight to the provider you chose, and that provider’s own privacy terms apply.

Nothing loads from other companies. The text and fonts come from Octopus itself. Wiktionary opens only when you choose it. There are no cookies, no advertising trackers and no analytics.

To remove everything, clear this site’s data in your browser’s settings. Anyone who uses this browser can open your library, so on a shared computer, clear it when you are done.

If you sign in

Signing in is optional, and only keeps your library in step across your devices.

Your library is encrypted in your browser before it is sent, with a library passphrase only you know. The server stores it but cannot read it: not your notes, not which verses you saved, not your verdicts. If you forget the passphrase, the synced copy cannot be recovered; the copy in each browser is untouched.

Octopus keeps only your email address, and uses it only to send sign-in links. The account and the encrypted library are stored with Supabase, in the United States.

If you sign in with Google, Google will know that you signed in to Octopus. Google also sends your name and picture; Octopus deletes them as they arrive and keeps only your email. If you would rather Google not know, use the email link.

Your library is yours to take. My Library exports it to a file at any time, with or without an account. Deleting your account deletes everything the server holds for you, at once.

Sources

Arabic text: Tanzil Quran Text (Uthmani, Version 1.1), © Tanzil Project, reproduced verbatim. Source: tanzil.net.

Word morphology: Quranic Arabic Corpus, version 0.4, © 2011 Kais Dukes, GNU General Public License. Source: corpus.quran.com.

Word-by-word glosses: written by Octopus, one plain meaning per dictionary word, put together for each word from its grammar in the Quranic Arabic Corpus. Lane’s Lexicon is the reference. A gloss is a literal aid for reading the Arabic, not a translation.

Root meanings: Edward William Lane, An Arabic-English Lexicon (1863–1893), public domain. Octopus shows the first sense of each root’s entry, and the whole entry, with Lane’s page numbers. Text provided by Perseus Digital Library, with funding from The U.S. Department of Education and The Max Planck Society.

Translation by Talal Itani, ClearQuran.com, under CC BY-ND 4.0.

Translation by Rashad Khalifa, Quran: The Final Testament, shown with permission. It leaves out 9:128–129.

Copyrights of all translations belong to their respective authors.

Typefaces: Amiri Quran, Gentium Plus and Source Sans 3, SIL Open Font License 1.1.