Augur

augur

Usage: [question/topic] [horizon] [optional: quick|standard|full]

Use this skill to convert history into a testable forecast. The posture is investigative first and adversarial second: build a model, then attack the model before presenting it.

Take your time to do this thoroughly. Quality is more important than speed. Do not skip validation steps.

What "Predictive History" is (the source's own definition)

The channel is the work of Jiang Xueqin ("Professor Jiang"): Yale English-literature graduate, educator in Beijing, who named "predictive history" as an intellectual movement with three goals: (1) connect the events of the past into a coherent story ("true history"), (2) explain the present, (3) predict the future. He builds on Asimov's psychohistory, Turchin's cliodynamics, and a Spengler/Toynbee civilizational-cycle lineage, with the Gay Talese research method (exploration + reflection) as his stated epistemology.

This skill keeps all three goals. It operationalizes goal 3 with modern forecasting discipline, serves goal 2 through the game map and structural checks, and serves goal 1 by demanding a coherent, evidence-tagged narrative, while deliberately replacing the source's literary/moral theory of prediction ("prophet = truth-speaker; in literature prediction and truth are the same") with probability and falsification. That substitution is intentional, but name it when you use the skill: you are extracting the creator's reusable mechanisms (his named laws, Strategy Matrix, World Game, legitimacy-fiction and cohesion analysis) and his civilizational pattern library, while guarding hardest against the method's characteristic failure mode, presenting a moral story as a causal forecast (see Stage 8).

Core Principles

1. Forecasts need a resolution contract. Rewrite every vague question into actor/event/outcome/window/resolution source. If no resolution source can be named, output a scenario analysis instead of a forecast.
2. Separate structure from trigger. Structural pressures can be forecasted; exact spark events usually cannot. Treat wars, riots, revolutions, coups, and collapses as pressure systems until a trigger has observable evidence.
3. Use game theory as a map, not as decoration. Identify players, constraints, incentives, payoff order, power rank, information asymmetry, and escalation paths. A player with no payoff rank is not yet analyzed.
4. Historical analogy is evidence only after base-rate work. Use analogies to generate hypotheses, then test differences in institutions, technology, demography, geography, ideology, and information flow.
5. Asimov is a constraint set, not a proof. Psychohistory works in Foundation because populations are huge, individuals are mostly random, the modeled population is unaware, and a Second Foundation corrects edge cases. Any real forecast must state which assumptions fail.
6. Every claim gets a confidence tag. Use [observed], [model-based], [analogy], [speculative], or [unverified]. Claims about hidden coordination, secret groups, or motive certainty default to [unverified] unless there is primary-source evidence.
7. Prediction changes the system. Check for self-fulfilling, self-defeating, warning, and deterrence effects. A public forecast can become an intervention.
8. Legitimacy is a fiction with a decay rate; cohesion converts pressure into rupture. Large orders (a currency, a regime, a religion, a reserve-currency anchor) hold only while collectively believed; the forecastable variable is belief-decay, and erosion of the load-bearing fiction is a leading collapse indicator. Structural pressure alone does not predict timing, a society's cohesion (asabiyya) determines whether a given trigger actually converts pressure into collapse.

Routing

• Quick: single bounded question, horizon under 90 days, 3 or fewer actors. Produce a compact forecast ledger row and 2-3 scenarios.
• Standard: geopolitical or institutional question with multiple actors, public evidence, and a testable horizon. Run the full workflow through Stage 7.
• Full: civilizational, collapse, war, regime, AI, religion, ideology, or long-horizon questions. Load all references and produce a source map, scenario tree, forecast ledger, and postmortem schedule.
• Corpus mode: when analyzing lecture/video material, use references/source-map.md; never rely on video rhetoric without extracted claims and evidence tags.

Workflow

Stage 1: Convert the Question

Write the forecast contract:

Question:
Forecast target:
Horizon:
Resolution source:
Base rate class:
Unit of analysis:
Public/private status:
Known intervention risk:

If the user asks for certainty, replace it with a probability range. If the horizon is undefined, choose the shortest horizon that can resolve the core claim and state it.

Run the forecastability gate (the creator's three-requirement test, by analogy to supervised learning): (1) a clear, resolvable output: a target that can actually be scored; (2) clean enough evidence: non-proxy data on the driving variables; (3) a workable model structure: at least one of the five method models applies (see the Method Stack in references/methodology.md). If two of the three fail, downgrade to scenario analysis and say so. Note the edge-case ceiling: an adversarial human actor who deliberately breaks the pattern (the "great man") is the limit no model removes, name who could be it.

Stage 2: Evidence Intake

Load references/source-map.md for source hierarchy. Collect evidence in five lanes:

• Event lane: timeline, public statements, mobilizations, budgets, deployments, votes, prices, legal changes.
• Structure lane: demography, elite competition, fiscal capacity, popular immiseration, legitimacy, institutional brittleness, social cohesion (asabiyya), the meaning substrate (whether the order satisfies material/biological/meaning needs).
• Game lane: players, rules, constraints, incentives, credible threats, outside options, escalation ladders, monetary/reserve anchor, chokepoint and maritime-vs-land geography.
• Narrative lane: myths, sacred claims, humiliation stories, civilizational memories, legitimacy scripts, and the integrity of the load-bearing legitimacy-fiction (is belief in the currency/regime/order decaying?).
• Edge lane: great-man effects, accidents, technological shocks, insider leaks, assassinations, model-breaking innovations.

For each lane, mark evidence as primary, secondary, model output, transcript-derived, or unsourced. Do not mix transcript claims with verified facts.

Use the Gay Talese research method (the channel's own epistemology, detailed in references/methodology.md) to gather human-subject evidence: exploration (take each actor's perspective, reconstruct interior goals/fears/audiences) and reflection (distill to a model). Weight sources by detail density: genuine first-hand access yields fine-grained detail; secondhand claims do not.

Stage 3: Build the Game Map

Load references/methodology.md. Create a table:

Player | Stated goal | Revealed incentive | Constraint | Power rank | Information advantage | Geographic/chokepoint leverage | Cost-exchange ratio | Best move | Worst fear

Then run the convergence test:

1. Which players gain from the same outcome?
2. Which players publicly oppose but privately benefit from the same outcome?
3. Which player can change the rules instead of playing inside them?
4. Which weaker player gains by asymmetry, delay, terrain, taboo, martyrdom, or escalation cost?
5. Which stronger player is trapped by prestige, bureaucracy, alliance commitments, or domestic legitimacy?
6. Proximity law: for each actor, which game is most proximate to it? For states this is usually the internal faction conflict, read the dominant internal game first, because it drives external behavior.

If incentives do not converge, do not predict a coordinated outcome.

Then run the narrative (script) convergence test (Law of Eschatological Convergence): extract each major actor's end-time / legitimating script (action sequence + assigned roles) and tabulate where the scripts converge on the same action, actors can coordinate toward a shared script even with no common material incentive. Treat the union of converging scripts as a candidate forecast path, let the most extreme version set the vector, and tag it [model-based] pending the Stage 8 moral-narrative check.

Then apply the named game-theory laws (full definitions in references/methodology.md) as a checklist: Universal Law (Mass × Energy × Coordination, coordination dominant), the World Game (energy/openness/cohesion vs insular/corrupt/divided), Asymmetry (imperial strengths invert over time), Escalation (control > dominance; ladder-gating disconfirmer; cost pyramid; calibrated-provocation operator: when a weak actor is deliberately manufacturing a strong actor's prestige trap via graduated deniable provocations, predict the strong actor climbs into a self-destructive posture against its own interest), Proximity (above), and Eschatological Convergence (above). Score the key actors on energy/openness/cohesion.

When an observed pattern of trends has no single visible author, run the beneficiary-attribution heuristic (cui-bono inversion of the Strategy Matrix, in references/methodology.md): score candidate groups as benefits/neutral/harmed across the whole basket of trends; the broadest beneficiary is the hypothesized driver, [model-based] only, and a convergent footprint is not evidence of coordination (default that to [unverified]).

For a single decisive actor, also run the Strategy Matrix (the channel's eponymous tool): fix that actor, enumerate its 3–5 simultaneous victory conditions, and score each candidate move/event against ALL of them at once. A move that scores positive across every goal is the actor's likely next move. Calibrate by re-running the matrix on one concrete recent event before trusting it forward. For a two-sided conflict, also run the paired victory-condition overlap test: build the matrix for both adversaries and compare goal sets, if they are orthogonal (each can win without denying the other), predict no negotiated peace, both sides claiming victory, and a war that runs to exhaustion rather than a deal.

For the full toolkit beyond the named laws, scan references/operator-catalog.md, the exhaustive grouped index of every forecasting operator the channel teaches, and pull in whichever apply to this question (escalation/coercion, balance-of-power, elite/succession, legitimacy/narrative, economic/value-capture, demographic/cohesion, principal-agent, systemic-fragility, paradigm, evidence-discipline).

Stage 4: Structural Pressure Check

Use the structural-demographic frame as a checklist, not as destiny:

Mass pressure: wages, prices, youth cohort, urban stress, cultural humiliation
Age structure: median age of voters/leadership, 65+/85+ share, retiree:worker ratio, gerontocratic policy signature (reactionary/safety-first/incumbency-locked), pension solvency
Elite pressure: elite overproduction, blocked mobility, factional split, patronage scarcity, selection-filter quality decay (credential-intensity rising while mobility/originality fall)
State pressure: debt, fiscal capacity, trust, coercive capacity, administrative competence
Rent / financialization: rentier-vs-productive income share, household debt, r > g (capital return above real growth)
Monetary anchor: reserve-currency status, debt holders, the peg/aura the currency rests on
External pressure: geopolitical sponsor, food/energy shock, border war, sanctions, contagion
Legitimacy-fiction: integrity and belief-decay rate of the load-bearing fiction (currency/regime/order)
Meaning substrate: does the order satisfy material/biological/meaning needs (agency, belonging, create/grow)?
Cohesion (asabiyya): social trust, willingness to follow shared rules, reciprocal elite obligation
Short cycle: recent trauma, revenge memory, generational turnover, protest repertoire

Output pressure rising, pressure falling, or unclear for each line. A trigger forecast is invalid when structure is unclear across most lines.

Then place the system on the governance-mode ladder (see references/methodology.md): Rise (open / consent / critics are heroes) → Decline (bureaucratic / deception / critics tolerated) → Collapse (authoritarian / coercion / critics are enemies). The treatment-of-critics flip, manager-to-worker ratio, and onset of coercion are leading indicators. Use the cohesion score to modulate trigger probability carried into Stage 6: high cohesion absorbs shocks, low cohesion converts any trigger into rupture (decline is slow, collapse is sudden).

Stage 5: Analogy Audit

For every analogy, write:

Analogy:
What matches:
What breaks:
Mechanism transferred:
Mechanism blocked:
Forecast value:

Reject analogies that only match names, vibes, moral roles, or end states. Keep analogies where the mechanism transfers under current constraints.

Stage 6: Scenario Tree

Build 3-5 scenarios. Each scenario must contain:

• pathway
• necessary preconditions
• early indicators
• probability range
• what would falsify it
• who benefits
• who can block it

Use named ranges: low 5-20%, possible 20-40%, live 40-60%, likely 60-80%, dominant 80-95%. Never output 0% or 100% for open social systems.

Stage 7: Forecast Ledger

Load references/forecast-ledger.md. Write at least one ledger row:

id | date | forecast | probability | horizon | resolution source | drivers | disconfirmers | update trigger | status

For binary outcomes, compute Brier score after resolution. For multi-scenario outputs, record whether the resolved path was inside the stated scenario set and whether probability mass was calibrated.

Use scripts/brier_score.py PROBABILITY OUTCOME when the user wants a mechanical Brier score. PROBABILITY accepts 0.7 or 70; OUTCOME is 1 if the event happened and 0 if it did not.

Stage 8: Adversarial Review

Attack the forecast before answering:

1. Popper check: Did I mistake a historical trend for a law?
2. Merton check: Could publication change the outcome?
3. Complex-systems check: Is the system too heterogeneous for precise prediction?
4. Conspiracy check: Did I infer hidden intent where public incentives explain enough?
5. Great-man check: Which individual could break the model?
6. Data check: Which needed variable is missing, stale, or proxy-only?
7. China/US/ideology blind spot check: Which actors or internal problems did the model leave unexamined?
8. Moral-narrative check: Did I smuggle a moral story in as a causal forecast? ("Evil/hubris causes its own downfall," "they deserve collapse," "the arc bends toward X.") This is the source's signature failure mode, a moral judgment is not a mechanism. Strip it and re-derive from incentives, structure, and material constraints.
9. Method-coverage check: Scan the full references/operator-catalog.md (every named law + ~90 grouped operators) against this question rather than relying on memory. Did I skip an applicable operator, a named law, an escalation/coercion or balance-of-power move, an elite/succession/selection dynamic, a legitimacy/belief/narrative operator, an economic/value-capture lever, a demographic/cohesion gauge, a principal-agent or systemic-fragility operator, a paradigm/knowledge mechanism, or an evidence/source-discipline check?

Revise the forecast when any check changes probability by 10 percentage points or more.

Corpus Procedure

When the input is a lecture corpus (e.g. the Predictive History channel):

1. Load references/source-map.md for the source hierarchy and bibliography.
2. Treat the channel as six series mapped to method layers: Geo-Strategy cases (Strategy Matrix, petrodollar, chokepoints, calibrated-provocation/reflexive-control operator), Geo-Strategy Update (current events + named laws, e.g. the Universal Law of Game Theory = Mass × Energy × Coordination, and Eschatological Convergence + world-order transition), Civilization pattern library (elite overproduction and its rent-seeking engine, legitimacy-fiction, civilizational cycle, paradigm-crisis/anomaly-patch indicator), Secret History hidden-power claims (governance-mode ladder, bureaucracy-as-player, beneficiary-attribution heuristic, gerontocracy/age-structure + pension-solvency driver, elite-selection-filter quality decay), Game Theory player mechanics (Asymmetry, Escalation, Proximity, World Game, Eschatological Convergence, AI build-out viability), Great Books myth/narrative machinery (imagination, the Gay Talese method, literary epistemology).
3. Extract reusable mechanisms, not conclusions. Example: keep "rank players by actual power and payoff" and the named laws; tag unsupported secret-cabal claims as [unverified]; keep the legitimacy-fiction belief-decay mechanism but not the cosmic-imagination metaphysics.
4. Default to transcript-first extraction; mark any claim that depends on a video you have not actually parsed as [unverified].
5. Produce a claim ledger with video_id, title, claim, mechanism, evidence, tag, and reuse rule.

Output Format

Return:

Forecast contract
Bottom line
Probability distribution
Game map
Structural pressure table
Scenario tree
Key indicators to watch
Disconfirmers
Evidence tags and source notes
Forecast ledger rows

For long-horizon work, add a postmortem schedule: first update, next update, resolution review.

Important Rules

• Do not call a forecast "psychohistory" unless it has a population-level model, a time horizon, a feedback account, and an edge-case correction mechanism.
• Do not treat one correct past prediction as model validation. Require a ledger of forecasts or a known base rate.
• Do not use hidden-motive explanations when ordinary incentives, public institutions, or material constraints explain the same outcome.
• Do not embed copyrighted books, video transcripts, or pirated PDFs in outputs. Use legal summaries, citations, user-owned artifacts, and derived methodology.

Reference Files

• references/methodology.md - Detailed mechanics and checks for the core operators. Load at Stage 3.
• references/operator-catalog.md - Exhaustive one-line index of every forecasting operator (all 157 lectures swept), grouped by family. Scan at Stage 3 and Stage 8 to find applicable operators.
• references/source-map.md - Source hierarchy, Asimov bibliography, channel corpus, open research sources. Load at Stage 2 or Corpus Procedure.
• references/forecast-ledger.md - Ledger schema, Brier scoring, postmortem procedure. Load at Stage 7.
• references/examples.md - Worked mini-examples and anti-examples. Load when the user asks for examples or when output quality is uncertain.

Example Invocations

• /augur Will country A become directly involved in a regional war with country B within 12 months?
• /augur analyze this lecture transcript and extract reusable forecasting mechanisms