| Work / author / dates | Kitāb al-Maqālīd al-malakūtiyya (كتاب المقاليد الملكوتية), "The Keys of the Kingdom", seventy chapters called keys (iqlīd, pl. maqālīd) · Abū Yaʿqūb Isḥāq b. Aḥmad al-Sijistānī, Ismaili dāʿī and philosopher, d. after 361/971 |
| Part added | "The keys of negation": 15 of the 70 keys, in the book's order: IX, X, XII, XIII, XVIII, XIX, XXVII, XXVIII, XXXI, XXXII, XXXIII, XXXVII, XL, XLIX, LII (11,489 words). The selection and its reasons: docs/research/sijistani-scope.md §7. Not included: the other 55 keys; the editor's introduction, notes, glossary and indices. |
| Id / files | maqalid · data/sources/maqalid-base.json (built by scripts/recitals/maqalid_base.py), maqalid-corrections.json and maqalid-uncertain.json (the OCR corrections, by scripts/recitals/maqalid_correct.py), maqalid-meta.json, maqalid.json (vowelled), maqalid-analysis.json (word ids mq:<passage>:<token>), AI layer app/src/content/ai/maqalid.json |
| Reader route | /read/maqalid · about /read/maqalid/about |
| Kind | a new catalogue kind, Ismaili philosophy (ismaili). None of the existing kinds fits: "Philosophy of light" is the Ishrāqī and Ghazālian line, "Sufi handbooks" the manuals of the path. Al-Sijistānī is a philosopher of another school (the Ismaili Neoplatonism of the Command and the Intellect), and the kind leaves room for his Iftikhār, the Ithbāt al-nubuwwāt, al-Kirmānī and Nāṣir-i Khusraw. |
| Status | working edition from one digital text (an OCR), established 3 Oct 2026 by Claude · no page image or print compared · reviewed by a person: — |
1. Witnesses
| Siglum | Witness | Kind | Vowelled? | Licence | Location |
|---|---|---|---|---|---|
| O | OpenITI 0361AbuYacqubSijistani.Maqalid.Kraken210622021812-ara1: an uncorrected Kraken OCR (2021, model arabPers-WithDiffTypefaces) of the edition by Ismail K. Poonawala, Beirut: Dār al-Gharb al-Islāmī, 1432/2011 (Silsilat al-dirāsāt al-Fāṭimiyya; title page OpenITI ll. 82–83, preface dated 12 Sept 2010, ll. 197–199). 129,535 tokens, 16,240 lines; the author's text is ll. 1135–~9220, the rest is the editor's introduction, notes and indices. Page tags give the print's pages. | digital (OCR of a modern edition) | no | OpenITI corpus CC BY-NC-SA 4.0; the edition is © as an edition | GitHub OpenITI/0375AH, master, fetched 2 Oct 2026: the file |
| P | Poonawala's print (2011), the edition behind O. Not held. | critical ed. | partly | © | No open scan exists (searched 3 Oct 2026: archive.org full-text and metadata search for the title in Arabic and Latin script, and the web; it is in print, sold by the publisher and booksellers; catalogue records only: Stanford SearchWorks, Google Books, al-Kindi/IDEO). |
| — | Translations: none in the public domain; no complete English translation exists. Partial, in copyright: Paul E. Walker, Early Philosophical Shiism (CUP 1993), with passages of the Maqālīd [REPORTED], not held; Khalil Andani, "Shīʿī Ismāʿīlī Tawḥīd" (2023), held, translates three short passages (pp. 80, 81, 87) and paraphrases others (pp. 65–67, 70, 80–81, 86–88, 154). | © | Andani's three sentences are in the private study layer only, for §10b |
The manuscripts are known only through P's apparatus: the editor's sigla ز and ه (his two manuscripts, "the two copies"), and for key XXVIII al-Kirmānī's quotation in al-Risāla al-Muḍīʾa and a family manuscript (his note at O ll. 3693–3698, not reproduced). See his introduction for their description.
2. Base and why
O is the only digital text of the work, and P is the only edition. The Maqālīd is the open work of al-Sijistānī that states his apophatic doctrine most fully, in short self-titled chapters (scope §7); his Yanābīʿ, which the secondary literature usually quotes, exists only in copyrighted editions and translations.
Only the author's text is taken. The editor's apparatus is stripped by maqalid_base.py and never reproduced:
- His footnotes at the foot of each page (manuscript variants, his corrections, verse references, glosses: 448 lines) are found by a page rule: the run of note-like lines before each page tag (sigla ز/ه, "the two copies", "as we corrected it", sūra references), stopping at any line that carries a footnote number fused inside it (only the author's text does). All 448 dropped lines were listed and read; two notes the rule misses are skipped by line (O ll. 4564–4565), and his long note inside key XXVIII (ll. 3693–3699).
- His manuscript-folio markers «[47]» (55), the footnote numbers fused to words (291), OpenITI milestones (40), his brackets round Quran quotations (27) and round words he supplied (21, the words kept): 434 mechanical removals, each logged with its line in
maqalid-base.json→changes. - His notes are read privately by the correction step as evidence of what the print reads (§3). Publicly they are only cited ("the editor's note records the manuscripts' reading"), never quoted.
Selection (OpenITI lines; a page tag closes its page, so a line's page = the next tag's number):
| Key | Title (our English) | ¶ | OpenITI lines | pp. | words | corrections | questioned |
|---|---|---|---|---|---|---|---|
| IX | God is not a substance | 0–7 | 1927–2057 | 72–76 | 998 | 57 | 2 |
| X | Whoever worships God by denying attributes and limits has not worshipped Him truly | 8–18 | 2066–2221 | 77–82 | 1,382 | 74 | 1 |
| XII | The one of the numbers and the One that is not of the numbers | 19–22 | 2303–2371 | 86–88 | 610 | 38 | 3 |
| XIII | Likening God to things is a hidden denial of Him | 23–26 | 2379–2438 | 89–91 | 499 | 47 | 1 |
| XVIII | The Sūra of Sincerity agrees with the testimony of faith | 27–32 | 2842–2932 | 107–110 | 716 | 50 | 5 |
| XIX | The Command mediates between the Creator and His creation | 33–38 | 2940–3030 | 111–114 | 615 | 36 | 2 |
| XXVII | The word «lā», the particle of negation | 39–45 | 3570–3670 | 136–139 | 816 | 52 | 3 |
| XXVIII | The Command of God is beyond being created, being the cause of creation | 46–49 | 3684–3776 | 140–143 | 495 | 35 | 2 |
| XXXI | Being does not befit the originated, let alone the Originator | 50–53 | 4010–4078 | 152–154 | 499 | 47 | 3 |
| XXXII | The Intellect is the most submissive of creatures to its Originator | 54–56 | 4096–4153 | 155–157 | 416 | 31 | 3 |
| XXXIII | The soul passes from potency to act only as mediator between Nature and Intellect | 57–61 | 4160–4229 | 158–160 | 645 | 45 | 2 |
| XXXVII | When the soul leaves the body, does it lose any of its knowledge? | 62–64 | 4550–4613 | 173–175 | 456 | 35 | 1 |
| XL | The resurrection is of spirits, not of bodies decayed and scattered | 65–72 | 4799–4957 | 183–188 | 923 | 58 | 2 |
| XLIX | The motions of the soul are endless | 73–79 | 5853–5950 | 222–225 | 791 | 52 | 1 |
| LII | The difference between revelation (tanzīl) and interpretation (taʾwīl) | 80–95 | 6241–6497 | 236–244 | 1,628 | 114 | 9 |
| 96 | 11,489 | 771 | 40 |
Pages are by the milestone rule; only pp. 80–81 (key X) were cross-checked against a scholar's citation (Andani 2023 cites the double-negation argument at pp. 80–81, exactly where O ll. 2153–2198 fall).
3. Collation and OCR correction
No second witness: there is no other digital text, and no page image (§1). The OCR was corrected by AI from context, and every correction is marked "supplied (AI)".
- Error rate. The scope note estimated 4–8 words in 100 wrong (out-of-vocabulary rate 6.8% on key X; a hand count of 5/110 and 10/150). The corrections made here, 771 on 11,489 words (6.7%), fall in that range.
- Method (
scripts/recitals/maqalid_correct.py, 3 Oct 2026; Claude Opus, headless, one call per passage). The model was given each OpenITI line raw (with the fused digits and markers) and cleaned, and the editor's footnotes on those pages as private evidence. It was told to restore what the print has, not to improve the author's Arabic. Each correction gives the line, the OCR reading (from), the correction (to), a reason, a basis and a confidence. A validator checked that eachfromis an exact, unique string of that line; failures were stated back to the model once.maqalid_base.pyapplies the high- and medium-confidence ones; low-confidence ones are not applied and are listed as questioned (maqalid-uncertain.json, 40, below). - What was corrected (771): by basis, context 373, parallel (the same phrase elsewhere in the key) 171, grammar 116, Quran or formula 76, the editor's note 75; by confidence, high 584, medium 187. Typical: dots (ج/ح: «وحد» → وجد, «حزئية» → جزئية; ت/ث: «أتبته» → أثبته), dropped letters («الي» → التي, «أعي» → أعني, «عرفوه» → يعرفوه), stray letters («سالبتسان» → سالبتان), quotation marks read as digits or letters («6ل1 موصوف» → «لا موصوف»), hamza seats («فإذأ» → فإذا). Full list with reasons:
data/sources/maqalid-base.json→changes(type "OCR correction"). - Reviewed by Claude (3 Oct 2026): all 74 corrections of key X were read against the raw OCR and the sense; none was rejected. One correction (XVIII, l. 2906) was re-anchored by hand because an earlier correction on the same line had already fixed part of it. The other keys' corrections were checked by the validator only, and by spot-reading.
- Questioned, not applied (40; the OCR's reading stays in the text). Among them: IX l. 1974 «هي مقول» (sense wants هو); X l. 2152 «بين الاستشهاد» (dual?); XII l. 2311 «تشبيه» or تثنية; XII l. 2329 «المبزورات»/المبروزات (al-Sijistānī's term for what is "sown" in the Intellect; left as O); XII l. 2359 «تلزم الخالق» or الخلق; XIII l. 2435 «فض» or نص; XVIII l. 2885 «لقسمعط» (garbled; probably لقسمة); XIX l. 3005 «وفايتة» (probably ونهايته); XXVIII l. 3686 «مخلوقة»; XXXI l. 4035 «يستبيع» (يستتبع?); XXXII l. 4104 «استقسه» (استمسك?); XXXIII l. 4220 «دكومة» (garbled); XXXVII l. 4553 «متجنبة». These are the first places a person with the print should check.
- What this cannot catch: an OCR error that is itself a good word and fits the sense (e.g. «إذا» for إذ), and any place where the print itself is faulty. The text is the print as best restored, not a critical text.
4. Decisions
No reading of the edition was knowingly changed: every correction aims at the print. Where the model judged that the print itself might differ from the manuscripts, it left the text alone and said so in its notes (the private correction log). Structure: each key starts a passage with its heading and title (the heading «الإقليد …» and the title as the edition prints them, corrected like the text). Very long paragraphs were cut at the sentence-final line nearest their middle; segmentation is fixed on the text before correction, so passage numbers do not move when corrections are added.
5. Quotations verified
22 Quran quotations replaced by the verbatim Tanzil Uthmani text (check_quotes.py maqalid: 22 verified, 0 failures). They were found from the verse references in the editor's notes (a fact we use; the notes are not reproduced): for each reference the quotation is located in the key by its dotless consonantal skeleton (robust to the OCR's dot errors), at least three words; a و/ف the author adds stays outside the quotation. Two references the skeleton matched wrongly in a first run (Q 23:71 at the wrong place, Q 42:13) were caught by reading the list, and the matcher tightened. Allusions without a reference are left to scripts/quran_refs.py.
6. Vowelling
O is unvowelled. Full vowelling by AI (scripts/recitals/prepare.py maqalid, Claude Opus), letters verified unchanged; Quran quotations restored verbatim afterwards (restore_quran.py). No vowelled edition compared. Ambiguous words are listed per passage in maqalid.json → ambiguous. Terms of art to watch: al-mubdiʿ (the Originator) and al-mubdaʿ (the originated) are one skeleton; the key titles of XXXI («لا يليق بالمبدَع فضلا عن المبدِع») turn on it.
7. Segmentation and word IDs
96 passages (our own, ~80–150 words, whole OpenITI paragraphs; the longest 257). Word ids mq:<passage>:<token>. Each passage records its OpenITI lines (GitHub link) and its print page(s).
8. Analysis layer
CAMeL Tools via scripts/fusus/analyze_ch.py maqalid.json maqalid-analysis.json mq.
9. Commentaries aligned
None. No commentary on the Maqālīd is known to exist. Al-Kirmānī's al-Risāla al-Muḍīʾa quotes and answers key XXVIII (the editor's note); it is not open.
10. Translation
AI English (Claude Opus, scripts/recitals/translate.py maqalid opus), badged, with the glossary in maqalid-meta.json (al-amr the Command, al-mubdiʿ the Originator, tashbīh likening, taʿṭīl denial, tajrīd isolating, ays/lays kept in transliteration where the wordplay matters). Told not to imitate Walker, Andani, Landolt or Corbin.
10b. Translation triangulation
See the change log for the run. The only translations held are Andani's three sentences (pp. 80, 81, 87, ¶15, ¶17, ¶21), copyrighted, private. Andani also argues a term divergence worth knowing: that Walker and Madelung read maḥdūd / ghayr maḥdūd as "having / not having a definition", where in the Maqālīd it means bounded (corporeal) and unbounded (spiritual) existents (his n. 11, citing pp. 95–96, a key outside this selection). Our glossary follows the "bounded" reading.
11. Resonances (PROTOCOLS §16)
scripts/resonances.py whole --only maqalid. Two connections are set by this record, not by the model:
- Plotinus, Ennead VI.9 (§§3–6; in the reader,
/read/plotinus-6-9): a parallel (resonance, not influence). The One known by negation and by presence, beyond the One of number; the soul strips every form to meet it. Whether VI.9 is among the treatises paraphrased in the Arabic Plotinus is not established here [UNVERIFIED]: do not claim that al-Sijistānī read VI.9. - The Arabic Plotinus (the Theology of Aristotle, paraphrasing parts of Enneads IV–VI): a documented lineage at the general level, as the editor states it: Poonawala's introduction (O ll. 143–150) says the Theology is drawn from Enneads IV–VI and that al-Sijistānī drew on it "to a great extent", with the Pure Good and the pseudo-Plutarchan Placita; his notes compare keys XII and XIX with Badawī's Aflūṭīn ʿinda l-ʿArab (for XIX, pp. 114–120) [REPORTED: editor; read on O, 3 Oct 2026].
12. Rights and private layers
Public: the author's tenth-century text as OpenITI provides it (CC BY-NC-SA 4.0), our corrections and our English. Not public: Poonawala's introduction, notes, variants and indices (never reproduced; cited by line only); Andani's translations, and the correction log with the notes as given to the model, are kept in the private study layer.
13. Open questions / to review
- Check the 40 questioned readings and a sample of the 771 corrections against the print (pp. 72–244). Until then every corrected word is "supplied (AI)".
- The print's own hamza and orthography in the key headings (O has both الاقليد and الإقليد).
- Which treatises of Enneads IV–VI the Arabic Plotinus contains, before saying anything about VI.9 beyond a parallel.
- Walker, Early Philosophical Shiism (1993): acquire, to triangulate the passages he translates (private layer).
14. Change log
- 2026-10-03: record created. Base (
maqalid_base.py): 15 keys, 96 passages, 11,489 words; 448 note lines and 434 apparatus marks removed; 771 OCR corrections applied, 40 questioned; 22 Quran quotations restored. Key X added to the heart-sight path (Plotinus VI.9.4 stop) as a further passage.