Doctrine X. Doctrine 10: Lineage Mining as Methodology — How to Walk the Recursive Source

Doctrine 07 named lineage mining as the principle: own the recursive source walk; the durable intellectual position is the same shape as the durable commercial position. The principle is correct and unfalsifiable in its general form. A reader who agrees with Doctrine 07 still does not know what to do on Monday morning.

This essay specifies the methodology. The walk has four stages, three working tests that the reader can apply to their own thinking, three anti-patterns the discipline exists to prevent, an instrumentation layer (the codex-tool) that makes the walk auditable, and a worked example — the Mercantile Thesis V2's actual lineage walk, with the deliberate omissions named so the discipline's edges are inspectable.

The methodology is not novel as practice. Senra has been demonstrating it on the Founders podcast since 2017; serious academics have practiced it for centuries; long-form journalism's better practitioners do it routinely. What is novel is publishing the walk as part of the canon — making the lineage map a first-class artifact rather than a private pre-write that disappears into the essay's footnotes. The codex-tool is what makes the published version operationally tractable.

I. The Four-Stage Walk

A lineage walk for any intellectual claim has four stages. Each stage names what the previous stage cited; the walk terminates when stages stop adding new sources or when the writer hits a primary source.

Stage 0: the canonical work. The work that the writer's claim is most directly downstream from. For the Mercantile Thesis V2's macro-cycle framing, Stage 0 is Ray Dalio's Principles for Dealing with the Changing World Order (2021). For the V2's sovereignty-and-AI framing, Stage 0 includes Aschenbrenner's Situational Awareness (2024). For the merchant-principle structural claim, Stage 0 is the Lineage canon itself (the 41-essay biographical study running from Mansa Musa to Jorge Paulo Lemann).

Stage 0 is what the average reader will recognize. It is the "this is what I read recently" layer. It is also where most thinking stops, and where the canon's competitors stop.

Stage 1: the work Stage 0 cited. What did Dalio read? Adam Tooze, William Strauss & Neil Howe, Niall Ferguson, Charles Kindleberger, Carroll Quigley, Will and Ariel Durant, the History of the Decline and Fall of the Roman Empire. What did Aschenbrenner read? Ilya Sutskever's compute-trajectory talks, the OpenAI scaling-laws papers, the GPT-3 / GPT-4 systems papers, Patel's SemiAnalysis hardware-economics work, the AGI-timelines literature from MIRI / Open Phil / METR. What did the Lineage canon's authors read? Senra's Founders podcast for the merchant-biography reading discipline; Felipe Fernández-Armesto's Civilizations for the cross-civilizational frame; the standard biographical references for each subject.

Stage 1 is where original work begins. A writer who has only read Stage 0 is rebroadcasting; a writer who has read Stage 1 can synthesize across multiple Stage-0 sources because they understand the substrate the Stage-0 sources share.

Stage 2: the work Stage 1 cited. What did Tooze read? Karl Polanyi, Eric Hobsbawm, the German economic-history school. What did Polanyi read? Marx, Maine, the political-economy tradition. What did Sutskever's compute-trajectory talks cite? Hans Moravec on biological-substrate compute, the canonical neural-network history (Hinton, LeCun, Bengio), the AI-winter retrospectives. What did Senra read? The biographies he cites in each Founders episode — and those biographers' biographies. Walter Isaacson's sources for Einstein and Steve Jobs and Leonardo.

Stage 2 is where the writer encounters the substrate of the substrate. By this stage the writer is reading 80-year-old or 200-year-old works, often in translation, often by authors whose cultural moment is no longer the writer's own. The reading is harder. The yield is also higher: Stage 2 is where the writer encounters the load-bearing intellectual frameworks that Stage 1 took as background and Stage 0 didn't even know it was assuming.

Stage 3+: continue until the well. A walk that terminates at Stage 2 is shallow; a walk that terminates at Stage 5 is deep. The right depth depends on the claim. A claim about the structure of late-Republican Roman commerce can walk Stage 3 to Mommsen, Stage 4 to Mommsen's primary sources (Plutarch, Cicero, the Roman legal-text tradition), and Stage 5 to the archaeological record. A claim about contemporary AI economics can walk Stage 3 to Solow / Schumpeter / Galbraith on technology-and-capital substitution, Stage 4 to the foundational political-economy tradition (Smith, Mill, Ricardo), and Stage 5 to the philosophical arguments that political economy was originally a response to.

The walk terminates when the writer stops finding new sources that meaningfully change the reading of Stage 0. That is the well. A writer who has hit the well for a particular claim can defend the claim against any Stage 0–Stage 2 challenge because the writer has already considered the Stage 3+ context the challenger doesn't know about.

II. Three Working Tests

The methodology is operationally testable. Three questions the writer can ask themselves before publishing any load-bearing claim:

The Senra test. Can you name what your favorite essayist read that shaped the essay you're about to ship a response to? If yes, you've done the Stage 1 walk. If no, your response is a Stage 0 response — you're arguing with the surface, not with the substrate that produced the surface.

The Senra test catches a specific failure mode: the writer who reads Tyler Cowen's blog post on industrial policy and writes a counter-argument to the blog post itself, when Cowen's actual position is informed by his reading of three Stage-1 sources the writer hasn't engaged. The counter-argument is correct about the blog post and wrong about the position. A response that engages the substrate (the Stage-1 sources Cowen is drawing on) is harder to write and more durable.

The Buffett test. Can you name what Charlie Munger read that shaped the way Munger argued? If yes, you understand Buffett's reading better than Buffett's biographers do. The Munger reading list (Cialdini, Galbraith, Bertrand Russell, Darwin, Franklin, the cross-disciplinary "lattice of mental models" canon) is itself the Stage-1 substrate behind the Berkshire Hathaway investment philosophy. A reader who has only read Buffett's letters is at Stage 0; a reader who has read Munger's reading is at Stage 1; a reader who has read Cialdini and Galbraith and Russell on Munger's terms is at Stage 2.

The Buffett test catches the failure mode of biographer-distance. The biographer's read of the subject is already a Stage-1 synthesis. Reading the biography puts you at Stage 1 of the biographer's reading, which is Stage 2 of the subject's. To stand at Stage 2 of the subject — where Buffett actually stood — you have to read what the subject read, not what the biographer read about the subject.

The merchant test. Who taught the merchant you are studying? Who taught that teacher? Mansa Musa was taught by the Mali court tradition that synthesized the Soninke, Berber, and Arab commercial inheritances; the Mali court tradition was taught by the Ghana Empire's earlier integration of trans-Saharan trade; that integration was shaped by the Aksumite and Phoenician commercial substrates that preceded it. A Lineage essay on Mansa Musa that stops at Stage 0 (Mansa Musa's own reign and decisions) is missing the load-bearing context. A Lineage essay that walks Stage 2 (the Soninke / Berber / Arab synthesis) and Stage 3 (the Aksumite / Phoenician substrate) can engage Mansa Musa as a node in a flow rather than as a singular figure, which is the merchant-principle reading the canon's frame requires.

The merchant test catches the failure mode of singular-figure history. Most popular biography treats the subject as a self-made figure who arose from background to alter their world. The merchant lens demands the opposite reading: every figure inherits a flow, modifies it, and passes it on. The walk surfaces the inheritance.

III. Three Anti-Patterns

The methodology has three failure modes that look like the discipline. The discipline rejects them.

Lineage shopping. A writer surveys the available Stage 1 / Stage 2 sources and selects the ones that flatter the writer's existing position. The walk reads as deep but actually only sampled the predecessors who happen to confirm the writer's thesis. The discipline rejects this because the walk's purpose is to surface the load-bearing substrate, not to assemble a pleasing-looking citation list. A walk that did not encounter at least one source that complicated the writer's position is not a walk; it is a confirmation tour.

The test for lineage-shopping: can the writer name a Stage 1 or Stage 2 source whose reading of the underlying problem disagrees with the writer's? The Mercantile Thesis V2's macro-cycle framing draws on Dalio (Stage 0); Stage 1 includes Tooze, whose Crashed reading of the post-2008 financial-system response disagrees with Dalio's frame on the role of central-bank balance-sheet expansion. A writer who has walked Stage 1 honestly should be able to name the disagreement and explain why they sided with Dalio anyway. A writer who cannot is either lineage-shopping or has not actually walked Stage 1.

Stage-0 capture. A writer reads Stage 0, treats it as the substrate, and never goes deeper. The writer's claims sound like Stage 0 because they ARE Stage 0 — paraphrased, recombined, occasionally extended at the edges, but not synthesized across the Stage-1 substrate that Stage 0 itself was synthesizing.

Stage-0 capture is the dominant failure mode in the contemporary essay-blog ecosystem. The economic incentive structure rewards rapid response to recently-published Stage 0 work; the time required to walk Stage 1 properly does not fit the publication cadence. A writer who is shipping weekly cannot walk Stage 1 weekly. The discipline therefore demands either a slower publication cadence or an explicit commitment to walk Stage 1 on the load-bearing claims even if the surface claims stay Stage 0.

The Mercantile Thesis V2 is an example of the demand: the V2's claim about AI 2026 is informed by a Stage 1 walk into the political-economy substrate of Dalio (Tooze, Polanyi, Kindleberger), and a Stage 1 walk into the AI-economics substrate of Aschenbrenner (Sutskever, Patel, METR). The walks are not visible in the V2's prose because V2 is a public-facing flagship optimized for readability; the walks are visible in the codex notes and in the Lineage canon's structural choices (which figures got profiles, which clusters got named, which counter-examples got first-class status). The walks are real even when they are not on the surface.

The "and the field" fallacy. A writer gestures at "the literature" or "the field" or "modern scholarship" without naming individuals. This reads as comprehensive and is functionally a hedge — the writer is claiming the support of a body of work without committing to any specific reader of that work. The discipline rejects this because the field doesn't argue; individuals argue. A writer who can't name the specific scholars whose work supports the claim is either bluffing or hasn't done the walk.

The fix is to name specific names with specific positions. "Tooze on the post-2008 response" is grade-honest; "the political-economy literature" is not. "Aschenbrenner's compute-trajectory framework" is grade-honest; "the AI-timelines community" is not. "Senra on the recursive source walk" is grade-honest; "the founder-biography podcast genre" is not. The discipline trades the rhetorical comfort of "the field" for the auditable specificity of named individuals.

IV. The Codex Substrate as Instrumentation

A lineage walk is too valuable to live only in the writer's head. The QM canon ships its walks through the codex, a Zettelkasten-style atomic-note substrate at ~/codex/ whose notes are linked by Obsidian-style [[wiki-links]]. Each note is one stage of one walk; the wiki-link graph IS the lineage map.

The codex-tool — a Zig CLI providing analytical commands over the codex graph — instruments the walk. Three commands directly support lineage-mining methodology:

codex source-priority ranks sources by the number of citing essays. The highest-priority sources are the ones the canon leans on most heavily, which means they are the ones the writer needs to have walked deepest. If a source ranks high on source-priority but the writer has only done the Stage 0 walk, the discipline says: deepen that walk before shipping the next essay that cites it.

codex archive-priority ranks archives by the number of citing essays. Archives are not active sources — they are repositories the writer can return to. The Edison Papers archive at Rutgers is an archive in this sense; the Hagley Museum's Westinghouse records are an archive. High-priority archives are the ones the canon will need to go back to repeatedly; the discipline says invest in the access-and-tooling layer for those archives so future walks are faster.

codex unread surfaces sources currently in the queued state — sources the writer has identified as load-bearing but has not yet walked. Sorted by citation count from the canon's existing essays (the highest-leverage queued sources first). This is the walk's worklist. A writer who runs codex unread and sees that Polanyi is the top queued source for three pending Lineage essays now has an explicit signal that walking Polanyi to Stage 2 will unblock three downstream artifacts.

The codex-tool's other commands (stats, links, orphans, dead-ends, validate, essay-graph, density, archetype-coverage, period-coverage, quality, voice-drift, combine-candidates) provide the broader analytical surface — a writer can audit the canon for orphan notes (sources with no citations, candidates for deletion or activation), dead-end notes (sources cited but never linked further into the graph, candidates for deepening the walk), or for missing cross-links between essays that should be connected.

The codex + codex-tool together make the lineage walk operational. A writer who runs codex source-priority weekly sees the canon's current substrate dependencies; a writer who runs codex unread weekly sees the next-highest-leverage walk to do. The methodology is no longer a vague injunction to "go deeper"; it is a specific weekly review of named sources sorted by canonical leverage.

V. Worked Example: The Mercantile Thesis V2's Lineage Walk

The Mercantile Thesis V2 cites three Stage 0 works directly:

• Ray Dalio, Principles for Dealing with the Changing World Order (2021) — for the macro-cycle frame
• Leopold Aschenbrenner, Situational Awareness (2024) — for the AI-compute trajectory
• Ben Thompson, "Aggregation Theory" (Stratechery, 2015) — for the platform-economics frame the merchant lens generalizes

Each Stage 0 cite is paired with a Stage 1 walk in the codex. The walk for Dalio: Tooze (Crashed, The Deluge, Wages of Destruction); Hobsbawm (The Age of Capital, The Age of Empire); Quigley (Tragedy and Hope); Strauss & Howe (The Fourth Turning); Kindleberger (The World in Depression). The walk for Aschenbrenner: Sutskever's 2015–2024 compute-trajectory talks; the OpenAI scaling-laws paper (Kaplan et al. 2020); SemiAnalysis (Patel) on hardware-economics; METR's evaluations work on agent autonomy; Patrick McKenzie on technology-supply-chain economics. The walk for Thompson: the original Aggregation Theory series at Stratechery (2015–2018); Joe Weisenthal's macro-takes adjacent reading; Albert Wenger's World After Capital for the platform-economics philosophical substrate.

The V2 also draws on the Lineage canon — 41 biographical merchant studies — as its own internal Stage 0 substrate for the merchant-principle structural claim. Each Lineage essay's Stage 1 walk is documented in its sources block; the cumulative Stage 1 walk across all 41 Lineage essays is the substrate the merchant-lens claim rests on.

What the V2 deliberately did NOT cite, with the reasoning:

• The post-2020 EU state-aid economic literature. Relevant to Bet 2's claim about wrapper-margin compression in EU markets, but the V2 was already at the high end of its target word count and the EU-market scope-limit was handled with a single-paragraph compression. A future Doctrine essay specifically on the EU AI Act's market-structure implications would walk this literature properly.
• The platform-economics labor literature (Vili Lehdonvirta, Mary L. Gray, the gig-economy academic tradition). Relevant to the wrapper-cluster's labor structure but the V2's frame is capital-and-substrate, not labor. The omission is honest; a future Lineage or Anti-Edison essay on AI-labor structure would walk this literature.
• The AI-safety alignment literature (MIRI, Anthropic Frontier Red Team, Apollo Research, METR's safety work specifically). The V2 brushes against AI-policy concerns but stays on the merchant-economics frame; a Doctrine essay on the alignment-meets-merchant-lens intersection would walk this.

The deliberate omissions are themselves part of the methodology. A walk that pretends to comprehensive coverage when it has actually scoped-down is dishonest. The V2's scope-limit is named in its body, and the omitted-literature list is auditable here.

VI. The Walk as Compounding Asset

A single lineage walk is a one-time investment. The canon-wide walk is a compounding asset.

Each Stage 1 walk the writer does on a load-bearing source informs the writer's reading of every other essay that cites that source. A walk through Tooze's Crashed informs the V2's macro-cycle claim AND informs the next Anti-Edison essay's reading of the post-2008 utility-sector regulatory regime AND informs the next Lineage essay on Henry Ford's reading of the early-20th-century American political economy. The walk amortizes across every downstream artifact.

This is what makes lineage mining the bottleneck for thinking work, in Doctrine 07's frame. A writer who is at Stage 0 across the board can ship one essay per week per topic; a writer who is at Stage 2 across the board can ship one essay per week that draws on six topics simultaneously. The compounding effect on per-essay synthesis density is large.

The codex-tool is the instrument that makes the compounding visible. codex source-priority reveals which Stage-1 walks would unlock the most downstream essays; codex unread reveals the next highest-leverage walk; codex stats reveals whether the canon is accumulating walks or stagnating at Stage 0.

The discipline's working test, finally: a writer who applies lineage-mining methodology should see their per-essay synthesis density rise over time, not their per-essay output rise. The methodology is not a productivity hack. It is an opposite-of-productivity hack — it makes each essay take longer and produce more. A canon practicing the discipline correctly will look slower from the outside (fewer essays per week) and denser from the inside (each essay drawing on three or four years of accumulated walks). That is the trade.

Doctrine 07 named lineage mining as the principle. This essay specified the methodology — the four-stage walk, the three working tests, the three anti-patterns, the codex-tool instrumentation, the worked example of the Mercantile Thesis V2's actual walk. The discipline is operational. A reader can now do the walk on their own canon-in-progress; a future canon-tool iteration can automate parts of it.

The methodology's edge: most contemporary essay writing is Stage 0 capture. A writer who walks Stage 2 routinely will be shipping a different category of work. That is the position worth holding.

Sources

Foundational:

• Doctrine 07 — Lineage Mining: Rubin Meets Grossman-Stiglitz, the principle this essay extends into methodology.
• David Senra, the Founders podcast (2017–present), the canonical contemporary practice of recursive source walks applied to merchant biographies.
• Rick Rubin, The Creative Act: A Way of Being (2023), the source-of-source obligation that Doctrine 07 names.
• Sanford J. Grossman and Joseph E. Stiglitz, "On the Impossibility of Informationally Efficient Markets," American Economic Review 70:3 (1980), the formal proof that information rents accrue to the sources others have not yet walked.

Adjacent:

• Charlie Munger, Poor Charlie's Almanack (2005, expanded 2023), the worked example of the Buffett test from the inside.
• Niklas Luhmann, the Zettelkasten methodology (the German original; How to Take Smart Notes by Sönke Ahrens, 2017, is the contemporary English-language reference). The codex-as-substrate is a Luhmann-style Zettelkasten.

Cross-references in the canon:

• The Mercantile Thesis V2 — the worked example whose Stage 0 cites and Stage 1 walks this essay's §V audits.
• Doctrine 08 — Capability-Graded Doctrine — the grading discipline that scores the walks. A walk that produced an over-confident grade reveals that the walk was shallower than it appeared.
• Doctrine 09 — The Dual-Receipt System — the procedure-receipt for any lineage claim is the published walk itself; the codex graph is the auditable substrate.
• The Lineage canon — 41 biographical merchant studies as of 2026-05, each one a worked example of the merchant-test walk applied to a single subject.

Footnotes