The Butlerian Jihad Reconstructed: Forty-Seven Campaign Records, 201–108 BG, and the Case for a War Among Machine-Augmented Human Factions

1. Introduction

The Butlerian Jihad, fought between 201 and 108 BG, ended some ten thousand years before the present Imperial vantage, and almost everything later ages believe about it arrives through intermediaries. Two accounts of its character circulate. The first is old and terse. Bene Gesserit teaching, as the sisterhood has occasionally let it be known outside its schools, holds that humanity once surrendered its thinking to machines in the hope of being freed, and that the surrender only allowed other men with machines to enslave them. The Orange Catholic Bible, compiled after the war by the Commission of Ecumenical Translators (C.E.T.), turns the lesson into its central commandment against building a machine that imitates the human mind. Neither source describes a machine ruler. Both describe a hazard in the relation between people and their instruments.

A second account is later and far more detailed. A body of compilations that this paper calls the late-tradition Synchronized Worlds narrative describes an ordered empire of thinking machines under a single intelligence, against which a unified humanity rose in a war of liberation. It supplies battles and heroes the older sources lack and has long shaped Corrino court historiography (Oskander, 10226 AG). Its authority is what this paper contests.

Our question is which reading the surviving campaign evidence better accommodates. We call the first the human-faction reading: the war was fought chiefly among human polities that had built machine capability into their administration and armies, and anti-machine unity was a framing that hardened afterwards. We call the second the machine-sovereign reading. The distinction matters because it changes what the Butlerian prohibition was for.

2. Sources and Method

The core evidence is a series of 47 engagement records held by the Salusa Secundus Military Archives as series JH-47. The series is our designation for a group of files that archivists at Salusa Secundus assembled from older holdings; no record in it is an original of the Jihad period. Each file is a copy, and several are copies of copies. We therefore recorded two properties of every file before reading its content: a dating grade and a copying date.

Dating grades follow the scheme one of us proposed for recopied pre-Guild material (Reyes-Okafor, 10238 AG). Grade A records can be placed within about three years; grade B records within about a decade; grade C records only within a phase of the war. Of the 47 files, 12 are grade A, 21 grade B and 14 grade C. Because grade C files cannot be sequenced finely, the reconstruction works with three broad phases: early (roughly 201–171 BG, 11 records), middle (roughly 170–140 BG, 19 records) and late (roughly 139–108 BG, 17 records). Sequencing within a phase is treated as uncertain throughout.

Copying date matters because the C.E.T. and the Imperial chanceries that followed it had strong reasons to present the war as humanity united against machines. Files whose earliest surviving copy was made after the prohibition's codification are classed as post-codification copies; 11 of the 47 fall into that class. Late-tradition material held in the Imperial Archives, Kaitain (Imperial Archives, Kaitain, 10228 AG) was read for comparison but was not counted as independent evidence, since its own provenance cannot be traced to any campaign record.

Each file was then coded by both authors, working separately, for a single question: does the record name human principals holding machine assets on both sides of the engagement, or does it describe machine forces with no human principal named? Records too damaged or too vague to answer were coded indeterminate. The two coders agreed on 41 of the 47 files at first reading (Cohen's kappa for the three categories about 0.77). Four of the six disagreements lay between two-sided and indeterminate, one between one-sided and indeterminate and one between two-sided and one-sided; after joint re-reading of the copying hands, three of the six were coded indeterminate, two as two-sided and one as one-sided.

Two further bodies of material were read alongside the engagement files. Casualty estimates survive for 23 of the 47 engagements. Five coalition compacts also survive, all in later Galach translation (Kesteven, 10241 AG).

3. Machine Capability Among Human Factions

Coding the 47 files gave three groups. Twenty-nine records name machine-holding human principals on both sides of the engagement; nine describe machine forces with no human principal named; nine are indeterminate. Of the 38 determinable records, 29 (76%, exact binomial 95% CI 60–89%) therefore fit the human-faction reading directly. Table 1, at the end of this section, gives the distribution by phase.

Two-sided engagements dominate every phase, though the pattern is weakest in the early phase, where records are also thinnest. The middle phase, which supplies the largest group of records, contains only two one-sided engagements against thirteen two-sided ones. We read this as the period in which machine capability had spread widely enough among rival polities that almost any serious engagement involved it on both sides.

Copying date sharpens the picture. Five of the nine one-sided records survive only in post-codification copies, against six of the 29 two-sided records (roughly 56% against 21%). The one-sided records, in other words, are disproportionately the ones most exposed to later editorial framing. The numbers are too small for inference, but the direction is what the human-faction reading predicts if a machine-sovereign overlay was added after the war.

Casualty figures add no independent support. Of the 23 engagements with surviving estimates, 17 are two-sided, four one-sided and two indeterminate, a composition much like that of the series as a whole (17 of 23 against 29 of 47). The figures are too heterogeneous in origin to model. Because the largest two-sided battles were also the ones that victorious houses had most reason to record, the concentration of casualty evidence in two-sided engagements may reflect reporting as much as destructiveness.

4. Flashpoints and Coalition Formation

The treatment of Corrin indicates how the later narrative may have been assembled. Standard Imperial chronology places the Battle of Corrin in 88 BG, some two decades after the Jihad had ended, as the contest between human houses that settled the ascendancy of the house from Salusa Secundus and gave House Corrino its name. The Synchronized Worlds compilations move the battle inside the war and recast it as the final blow against the machine sovereign. As the chronology implies, no file in series JH-47 corresponds to it. We read the relocation as a small, datable instance of the proposed overlay: a dynastic victory written into a war of liberation by the dynasty's own court historians.

A middle-phase sequence of eight engagements in the region the files call the Orl Marches supplies a second case. Three authorities that had divided a former provincial administration among themselves, each with its own machine establishment, fought one another there for most of a decade. A three-way contest among machine-holding authorities is hard to fit to a one-front war, and the machine-sovereign accounts omit it altogether.

Less secure is a third case, in which a cluster of late-phase files records a polity that kept machine-production assets while changing sides. We offer, as our own hypothesis, that such arrangements may be part of the pre-history of the later Richesian and Ixian tolerance; the files do not name any house that survives into the Imperial period.

Coalition compacts provide the most direct evidence of motive. All five surviving compacts justify alliance by house alignment, territorial guarantee and succession. Two carry preambles invoking a common struggle against machines, and in both the preamble is in a different copying hand from the operative clauses. On five cases this is suggestive and no more, but it fits the proposal that anti-machine unity was written over the compacts after the fact.

5. The Prohibition as Settlement

The commandment at the heart of the Orange Catholic Bible is canonically religious, and nothing in this paper disputes that status. Surviving C.E.T. session calendars (Commission of Ecumenical Translators, 10231 AG) show that the commandment was drafted as scripture and debated as scripture. The same calendars also show delegates from former belligerent houses negotiating its scope, which places the religious text inside a political settlement.

Under the human-faction reading, one plausible function of the settlement was to withdraw machine capability from every party at once. We state this as a hypothesis. The files record which sides held machine assets; outcomes survive too unevenly, and too often in the victors' words, to show whether those assets decided engagements. Bene Gesserit teaching records the same lesson in moral form (Bene Gesserit Archives, 10236 AG), and the Mentat Collegium's own histories present Mentat training as the human replacement for what the settlement removed (Ventiri, 10233 AG).

6. Discussion

Taken together, the campaign files, the flashpoints and the compacts are more economically explained by the human-faction reading. Most determinable engagements set machine-equipped human commands against each other; the records that show a machine force without a human principal are disproportionately post-codification copies; and the surviving alliances were made for dynastic reasons. None of this proves that no machine intelligence acted independently at any point in the war. It does mean that the older sources, which speak of men with machines subjugating other men, sit closer to the surviving evidence than the late tradition does.

If the war was fought among machine-augmented factions, the prohibition carried a dual function. It was, and remains, a religious commandment. It may also have served as a balance-of-power settlement, removing from every house at once an instrument that most belligerents in the determinable files had held; whether that instrument had been decisive in battle is a question the surviving records do not answer. A settlement reading of this kind fits recent work showing that the prohibition's durability rests on enforcement infrastructure working alongside religious conviction. It also fits technical evidence from the same period that thinking machines ran as federated, separately held installations with local authority rather than as a single synchronized intelligence; that architecture is what rival polities each owning machine establishments would produce.

The Ixian exception offers a present-day test. The inspection-protocol account holds that Ixian machinery is tolerated because inspectors can verify that it lacks autonomous cognition. We propose, as a hypothesis the present evidence cannot test, that this boundary coincides with the line the settlement was drawn to protect: permitted Ixian devices extend what a house can make without restoring the machine cognition that most belligerents in the determinable files held.

7. Limitations of the Record

Forty-seven engagements are a small, unrepresentative remnant of a near-century of war. The series survives because Salusa Secundus archivists chose to recopy it, and their principles of selection are unknown. Grade C files cannot be ordered within a phase, so any claim about sequence inside the three phases would outrun the evidence. The records are also not independent: eight come from the single Orl Marches sequence, and several others share a copying hand. The binomial interval treats each file as independent and so understates the uncertainty.

Coding depended on how each record names its principals, and a copyist who wished to erase a human commander could produce a one-sided record from a two-sided engagement; the reverse is harder to imagine but not impossible. Agreement between coders was substantial, but both coders share the same training and may share the same blind spots.

Casualty figures are the weakest material in the series. They come from victors, survive for fewer than half of the engagements, and differ in how dead were counted. They were used here only descriptively. Finally, no machine-side administrative record of the kind the late tradition describes has been located in the archives open to us; its absence is compatible with loss as well as with non-existence, and a future discovery could alter the balance of this reconstruction.