Skip to content

Changelog

All notable changes to this project will be documented here.

Format follows Keep a Changelog. Versions follow Semantic Versioning.

[Unreleased]

Fixed

Flagged cleanup after the docs pass (2026-09-12) - eoh dashboard no longer invents its accounts. The ledger, the Trust and the EOH totals were built from literals (destroyed = 0.85 × created, expenditure = 0.90 × earnings, total EOH = population × base × 1.5), so Conditions I and III passed by construction. EOH now comes from the pipeline and the ledger and Trust from one simulated period on the same frame; the certified competency share, which nothing tracks, is declared and printed (--certified-fraction). No status changed at any reference ε. - run_simulation reported a 2.10 multiplier for periods it minted at simulate_period's own default. - corridor_is_usable(cadence="continuous") still described the cadence as undecided; its note now reads the crossover and the analogue disagreement live. - API pages for prices, multipliers, trajectory, conditions, simulation, EOH generation and fulfilment corrected against the docstrings; knowledge EOH no longer claimed to become dominant (personal stays the largest domain), and personal registration no longer claimed to approach 1.0. - guf_framework.md describes the shipped per-parcel term and the ×100 use-rate scaling beside the NLSA specification.

Changed

The signed-off transparency pass reaches the front pages (2026-09-12) - overview.md, design_principles.md, three_economies.md and the arc waypoints now carry the positions adopted on 2026-08-15: ε → 1 as a limit, prices as floor prices, and currency that records and shows its work rather than measuring value. Design Principle 8 no longer says a green dashboard means the system works. - guf_framework.md gives the shipped fee (Ψ retired, monotone in ε) beside the NLSA bell in §4.1 and §4.4, and labels the worked example as the published NLSA form.

Added

The anchor comparison is published, and the docs are gated (2026-09-12) - docs/theory/anchor_comparison.md — What Anchors a Unit of Account, published as work in progress beside Prior Art. It quotes shapes where a figure is calibration and values only where a figure is structural; tests/test_anchor_page_figures.py reads the published page and checks every structural statement against utils/anchor_page_figures.py. - tests/test_doc_examples.py — every Python block in README.md and docs/ runs, every eoh_cli.py example parses, and every API signature heading names real parameters. 34 of 54 published examples and 95 signature headings had gone stale with nothing checking them. Template blocks and placeholder CLI lines are ratcheted. - scenarios.collective.collective_snapshot() accepts available_labor_eoh and rationing, and forwards the state's own deferred_ecological, knowledge_complexity and monitoring_capability. The institutional entry point could not previously verify fulfilment against labour at all, so every snapshot measured demand. Blast radius on a default state: zero.

The tag scheme: CHOSEN split into bounded / placeholder / normative (2026-08-09) - One tag was covering three epistemic states. It made the calibration set read as 83% guesswork while hiding which constants are the real debts, and it filed charter decisions under "awaiting measurement" — a category error, since no dataset settles what fraction of an estate should pass to heirs. - bounded (14) — picked inside a measured band; band: and errs: are required and gated, so "bounded" cannot be a free upgrade over placeholder. errs opens with HIGH / LOW / NEITHER / WITHHELD, the last a real epistemic state here. - placeholder (106) — nothing constrains it yet. The actionable debt. - normative (60) — a decision. Requires decided_by:, may carry precedent:, and is forbidden a resolves_by: — the forbidding is the point. - The eight CANONICAL_* arc constants became convention: an ideal-arc reference frame, whose own pointer already read "nothing, and by design". - The honest headline: debt is 52.4%, not 83.3% — grounded 49 (21.4%), bounded 14 (6.1%), placeholder 106 (46.3%), normative 60 (26.2%). eoh provenance check now prints this as a "Where the model stands" table via debt_summary(). - Of the 14 bounded picks 7 err LOW, and for DEP_RATE, THERMAL_F_GHG and the GOVERNANCE_IRR_* pair low is the unsafe direction — it flatters solvency, overstates the thermal allowance, and sets the assessment gate below the conventional κ bar. Each records that on its own line. - CDR_LABOR_HOURS_PER_TONNE → measured (Tier D): one plant's staffing disclosure is a measurement with n=1, and it now matches GUF_ECO_KAPPA_CARBON.

Two constants revalued (2026-08-09, author decisions) - RECAL_FOUNDING_LABOR_HOURS 1,000 → 666.67, and bound to PERSONAL_EOH_BASE via the new RECAL_FOUNDING_FRACTION = 2/3. The reprice had silently turned "two-thirds of the base" into the whole of it — a founder with nothing left to live on. Blast radius measured: the labour arm lengthens exactly 1.5× (t_labor 1.80 → 2.70 yr at ε=0, worst 2.85 across the arc), the three-channel exit invariant holds at every arc point, and the channel order is unchanged. TestFoundingLabourArm now anchors the mechanism — nothing had been exercising the labour arm's timing, which is why a 1.5× drift was invisible. - GUF_ECO_KAPPA_CARBON 2.750 → 0.6, adopting CDR_LABOR_HOURS_PER_TONNE as the better-sourced of two figures for the same physical quantity. Bound by test rather than expression (forward reference), so either side moving alone fails. Left open and pinned: whether CDR_GROSS_REMOVAL_FACTOR belongs in this path.

Provenance coverage closed and gated (2026-08-09) - utils/provenance.py + utils/provenance_cmd.py — inline provenance tag blocks in data.py are parsed from source, checked against the tag scheme, and rendered as the shipped audit CSV and the generated tables in docs/parameter_provenance.md. CLI: eoh provenance check | csv | table | doc. - hours_eoh/reference/data/constant_provenance.csv — generated, one row per data.py constant (value, units, tag, tier, form, block, resolves_by, note), so a public audit needs no Python. Packaged via the existing hours_eoh.reference data glob. - All 229 data.py constants now carry an inline tag block, up from an effective 3. 101 constants (44%) previously appeared nowhere in the provenance doc, 51 of them the entire GUF block. Nine doc tables were still on the retired binary Kind = Physics | Calibration. - tests/test_provenance.py — 69 tests. Parser behaviour is pinned against synthetic source; the coverage gate runs against the real data.py with no allowlist, and fails on an untagged constant, a stale or unmatched tag block, a retired or unknown tag, a missing epistemic pointer, a bounded constant with no band or no error direction, a normative constant with no decider or one claiming a resolves_by, missing units, a misapplied tier, a stale audit CSV, stale generated doc tables, or a block with no doc home. Each was verified to bite by breaking it and reverting. - tier (A–D) formalised as a sub-qualifier rather than a rival scheme, matching what the thermal layer already wrote; it applies only where there is a source to grade. convention promoted to a declared sub-label. The ad-hoc Physics-adjacent tag is retired, along with CHOSEN itself (see the split above).

Changed

Provenance doc: tables generated, prose hand-written - Every constant table in docs/parameter_provenance.md now sits inside a <!-- provenance:table --> region rendered from data.py. The prose around them — where the epistemic argument lives — is untouched and stays hand-written. New sections for Ground Use Fee, dashboard thresholds, capital lifecycle, TEH destruction, human capital and multiplier governance, none of which had one. - No calibration constant changed. Verified constant-by-constant against the previous commit: 228 compared, 0 value differences. - Only 2 of 229 constants are physics — applying the scheme's own demanding definition honestly leaves A_EARTH_M2 and SIGMA_SB and nothing else. The 190 that first became CHOSEN were then split (see above), because one tag covering three epistemic states was itself obscuring the picture.

Fixed

Four documentation drifts, found by the migration - KNOWLEDGE_EOH_BASE — doc said 490,107,421; the shipped value has been 381,962,855.27 since the ε_ref fixed-point re-anchor. - CARE_SIGMOID_DEFAULTS — doc said start_share 0.30 / inflection 0.55 against the shipped 0.05 / 0.45; it had never matched the code. - Membership min-hours thresholds printed as 750 / 1500 h/yr and per-capita personal EOH as 2,213 — both are products of PERSONAL_EOH_BASE and went stale at the 1,500 → 1,000 reprice. Now 500 / 1,000 and 1,475. A curated test pins these derived products, because no value-equality check can see a number restated in a sentence. - RECAL_FOUNDING_LABOR_HOURS was documented as "≈ 2/3 of PERSONAL_EOH_BASE" while the reprice had made it 100% of it. Decided and fixed — see "Two constants revalued" above.

Findings (reported, not resolved)

  • NLSA cites this framework's own document. All 51 GUF constants attribute to "NLSA Technical Manual TM-0042", whose own header reads "Based on NLSA from HOURSFramework". Equation numbers now appear only under form:, never resolves_by: — they establish an asserted form, not external evidence for a value.
  • GUF_ECO_KAPPA_CARBON (2.750 TEH/tCO₂e) and CDR_LABOR_HOURS_PER_TONNE (0.6 h/t) were the same quantity from two layers — a 4.58× disagreement, now RECONCILED (see above). Still open: DEP_RATE 0.045 vs FORMATION_DEPRECIATION_RATE 0.05, and CONTESTABILITY_CAPITAL_YIELD_RATE 0.10 vs the 0.20 derivable from RECAL_CAPITAL_OUTPUT_RATIO and FORMATION_DEPRECIATION_RATE.
  • Four constants are calibrated to a target and now say so: GUF_USE_*, DEFAULT_SEGMENTS, TRUST_BASE_TEH, CAPITAL_MACHINE_PROFILES — the _ECOLOGICAL_SPIKE_INTENSITY pattern the 2026-08-05 pass named but did not sweep for.
  • LEVY_SUFFICIENCY_WARN cannot fire on the shipped configuration: it warns below 2% guarantee coverage, and SUFF_LEVY_RATE delivers ≈2% at canonical defaults.
  • _ECOLOGICAL_SPIKE_INTENSITY lives in core/eoh_generation.py, not data.py, so the gate cannot see a constant the retag log covers — a standing violation of the no-anonymous-constants invariant.
  • Residual the gate does not close: free prose can still go stale. The curated derived-product test covers where the reprice drift actually hid; a narrative paragraph that goes stale in a way no field captures remains a human problem, and saying so is better than implying the test closes it.

Federation contestability closure (reconciliation §8.7 — research tier) - research/coasean.py Phase 4 — two-tier Trust: merge_collectives() and split_collective() boundary events with indivisible-reserve escheat (§8.7c) and TEH-conservation postconditions (§8.7d); simulate_federation(commons=True) tracks a federation commons funded by a levy tithe (COASEAN_COMMONS_TITHE, Italian Law 59/1992 precedent) plus consolidation escheats, and records per-collective χ each period (chi_min, chi_marginal_min, chi_worst_collective, chi_status_worst, commons_balance, commons_floor_coverage). τ counts both tiers. Defaults are byte-identical to Phase 3. - research/contestability.py — portable_endowment_federated() (two-tier P: the floor is federation-guaranteed and never vests; the dividend claim is per-collective), exit_value() (the floor crosses collective boundaries at par; only the capital account converts at the exchange rate), contestability_margin_federated(). Tenure is federation-wide (§8.7b): moving collectives never resets vesting. - research/membership.py — MembershipTerms + contestability_audit(): the §8.7e math/contract line. Admission cost adds to K_entry; exit notice, minimum hours, vesting length, and dividend retention are audited against MEMBERSHIP_* thresholds. - CLI: eoh coasean simulate gains --dynamics --g-priv --levy-rate --commons --commons-tithe --regime; new eoh contestability audit subcommand. - Honest adversarial findings at defaults (reported, not tuned): commons floor coverage is tiny at a 3% tithe, and consolidation escheat migrates trust from collective dividends to the commons across the arc, so the worst marginal χ worsens toward ε→1.

Contestability closure mechanisms (proposed §8.8 — research tier, pending author sign-off; see notes/contestability-closure-proposal.md) - Answers the Phase 4 adversarial findings. Root-cause diagnosis: (1) the commons was dividend-inert, so §8.7c escheat drained tenure-vested dividends into a fund paying no one; (2) the marginal (tenure-0) member holds no capital claim, so no levy schedule can close χ_marginal; (3) the static levy base ε·K·yield understates machine output ~12× at high ε (an ε=0-era capital calibration held fixed); (4) χ compares an annual endowment flow to a one-time founding stock. - M1 — universal commons dividend: portable_endowment_federated(..., commons_balance) pays the commons yield per capita UNVESTED (Alaska Permanent Fund precedent); escheat becomes a stabilizer — consolidation moves capital from tenure-gated collective dividends into the universal tier. - M2 — entry underwriting: entry_underwriting() + commons_seed_required(); the commons capitalizes new collectives' trusts (stays commonized, §8.7c). Combined invariant: exit_financeable ⇔ χ_marginal ≥ 1 OR entry_capacity ≥ 1. With a seed of ~1.8e7 TEH (~0.05% of the Trust base) the combined invariant holds at EVERY period of the canonical adversarial arc (asserted in tests) — while χ_marginal alone stays honestly CRIT at high ε (commons-financed, not self-financed exit). - M3 — physically-consistent levy base: machine_output_teh(ε) = ε·total_eoh(ε); levy_schedule_for_chi(levy_base="machine_output"). Growth steps of the schedule remain infeasible — M3 removes the calibration artifact, it does not manufacture feasibility. - simulate_federation(commons_dividend=True, commons_start=...) wires M1+M2 into the federation arc (commons_dividend_paid, entry_capacity, exit_financeable); contestability_audit(commons_dividend=, underwriting_policy=) audits terms against the combined invariant (CRIT → WARN waiver is opt-in; §8.7e defaults unchanged). New constants CONTESTABILITY_MIN_VIABLE_POPULATION (uncalibrated placeholder, documented as such) and CONTESTABILITY_UNDERWRITE_FRACTION. - CLI: coasean simulate --commons-dividend --commons-start; contestability levy --levy-base; contestability audit --commons-dividend --underwriting-policy. - All defaults preserve §8.7 behavior float-exact; 38 new tests (1595 total).

Recalibration prototype (proposed §8.9 — research tier; adopted-in-principle by the author 2026-07-26, formal reconciliation-doc edit pending) - research/recalibration.py — resolves the three §8.8 "honest remainders" at their causes. RC4 fix: the flow/stock χ is replaced by time-to-finance-exit (exit_financing(), invariant t_exit ≤ RECAL_EXIT_HORIZON_YEARS = one vesting period) and a genuine accumulating §8.7b capital account (capital_account_stock(), Mondragon internal-account precedent, zero-interest per Condition III). Open item 3 fix: capital_stock_epsilon() grows the stock with machine output (K(ε) = K₀ + ν·Y(ε), ν = Piketty's β ≈ 4) and the commons OWNS share φ(ε) of it (Meade social-dividend model), so τ = φ ≤ 1 by construction and dτ/dε ≥ 0 (Piketty inversion) is structural, not levy-contingent. The dividend is the commons' capital income net of share acquisition (commons_income_statement()) — 0 at ε=0, ≈1,873 TEH/person·yr at ε=0.99: funded by measured machine output, not a promise. - Self-financing dropped as the test (author decision): exit finance has three physical channels — own labor at low ε (the floor feeds the founders while they build), commons underwriting through the mid-arc trough (Caja Laboral / Marcora precedent), dividend savings at high ε. recalibrated_arc() shows the combined invariant holding at EVERY arc point in the adversarial regime, with the channel arcing labor → underwritten → self (asserted in tests). - Honest findings at defaults (reported, not tuned): share acquisition out of commons income is infeasible for ε ≲ 0.15 (the initial endowment φ₀·K₀ and human-era fiscal levies must carry the early arc); endogenous private-capital growth turns negative past ε ≈ 0.5 — the §8.2 commonization made visible instead of assumed away as a fixed g_priv. - trust_required_for_chi() and levy_schedule_for_chi() marked SUPERSEDED (retained unchanged as documented negative results: the trust-growth path cannot close bare χ — which is why underwriting exists). - New constants (all provenance-documented, placeholders flagged): RECAL_CAPITAL_OUTPUT_RATIO, RECAL_EPSILON_RATE_PER_YEAR, RECAL_FOUNDING_LABOR_HOURS, RECAL_EXIT_HORIZON_YEARS, RECAL_ACCOUNT_CREDIT_SHARE. - CLI: eoh contestability recal — the §8.9 arc table. 62 new tests (1657 total).

§8.9b — charter-formation doctrine (research tier; doctrine bundle agreed with the author 2026-07-26) - phi_policy on the recalibration module: "dilution" (new default doctrine), "target" (§8.9a purchase model, kept as the regression anchor), "escalated". - Charter formation (A3): the commons' share attaches to NEW capital at commissioning as a federation charter condition (resource-license / Georgist model) — formation_share_required() gives the per-ε charter share s(ε) (≈ 0.17 early, crossing 1 at ε ≈ 0.48). Nothing is purchased, so the §8.9a early-arc funding gap disappears and the dividend is the full φ·Y (self-financed exit from ε ≈ 0.30 vs 0.59 under §8.9a). formation_levy_rate() quantifies the compensated bridge (≈ 1% of labor-era output, sunsetting by ε ≈ 0.2). - Dilution ratchet (B2): private capital is NEVER sold down — it follows a running-max ratchet (rises while s ≤ 1, then flat). Honest cost, reported not tuned: φ caps at ≈ 0.66–0.68 by ε = 0.99 (target 0.99); the exit invariant still holds at every arc point with D ≈ 1,600 TEH/person·yr. - Escalation clause (B3): escalation_trigger() — adversarial regime observed AND (entry capacity < RECAL_ESCALATION_CAPACITY_FLOOR or invariant failing) → the charter takes all new formation (s = 1) and capital-estate escheat rises to RECAL_ESCALATION_ESTATE_SHARE. The trigger LATCHES in the arc; at canonical defaults it NEVER fires (asserted in tests; forced firing tested at a 40× founding cohort, lifting φ to ≈ 0.86 at ε = 0.99). - Generational conversion (B4): estate_conversion_flow() — capital estates escheat at RECAL_ESTATE_CAPITAL_ESCHEAT_SHARE (0.15 = ESTATE_LEVY_FRACTION: the D5 doctrine extended to capital, not a new rule). Honest finding: mortality speed is slow (half-life ≈ 69 yr even at full escheat) — φ → target is asymptotic over generations; §8.2's "φ must be ABLE to → 1" survives as an asymptotic capability, and the exit invariant never depends on it. - recalibrated_arc() path-integrates the capital split (TEH-conservation asserted); rows gain phi_target, cap_binding, s_required, escalation_active, private_capital_delta_per_year (the absolute-flow reporting fix — a rate on a vanishing base was theatrical). Open item, flagged: the charter share is an implicit tax on private capital formation; the investment-disincentive feedback on K(ε) is not simulated. - CLI: contestability recal --phi-policy --estate-escheat --min-viable-population; s_req/ΔKpriv columns, cap-binding and escalation markers. 79 new tests (1736 total).

§8.9c — formation feedback (research tier): who actually builds K(ε) - research/formation.py closes the investment-disincentive circularity §8.9b flagged open. Formation is FINANCED or it does not happen: private supply f(s) falls linearly as the charter share cuts the net return r_priv = (1−s)(1/ν−δ); the commons co-funds from net income per a priority policy; ε is DERIVED from the capital actually formed (aggregate inversion of the module's own physics; typed-capital integration via core/civilization.py is the follow-up). - Two corrections found in planning, now in the accounting: (1) the §8.9b funding hole — in the cap region s = 1 attracts zero private funding, so the commons pays for ALL formation there; (2) replacement — the commons must replace its own worn stock (δ·T_K ≈ a 20–24% dividend haircut); commons_income_statement(net_of_replacement=True) reports it (default False preserves published §8.9b gross figures). - Null anchor (charter share pinned 0): reproduces the canonical ~50-yr arc pace exactly — the baseline every feedback effect is measured against. - VERDICTS (honest, asserted in tests): share-first priority holds the canonical pace with ZERO delay — the feedback costs the arc nothing in time, but the dividend pays for it (D ≈ 113 vs static 302 at ε ≈ 0.4; self-financing onset moves from ε ≈ 0.30 to ε ≈ 0.86, underwriting carries the transition). Dividend-first never stalls but CRAWLS: ε ≈ 0.60 after 120 years, never completing. The exit invariant holds at every simulated year under BOTH priorities and even in the fiat counterfactual — capacity does not depend on the dividend. - THE CONDITION III FINDING: zero interest is the doctrine's structural ally. The incentive-compatible charter share is s* = 1 − r_full/r_gross = 0.50 at zero-interest returns vs ≈ 0.10 at fiat-like returns; in the fiat counterfactual the commons must drive the dividend to literally ZERO mid-arc to hold pace, where the zero-interest world never does. The framework's pieces reinforce each other, quantified. - New constants (provenance-documented, flagged): FORMATION_DEPRECIATION_RATE (from CAPITAL_MACHINE_PROFILES design lives), FORMATION_HURDLE_RATE_MIN, FORMATION_FULL_SUPPLY_RATE. CLI: contestability formation --priority --hurdle --full-supply --escalation --charter-share. 39 new tests (1775 total).

Collective land-inventory system (hours_eoh/land/) - collective.py — Standard parcel schema and batch GUF calculator for a collective land inventory. compute_collective_guf(parcels, epsilon, median_income) loops all parcels through ground_use_fee(), applies soil-health credits, review-cycle caps, and income-linked subsidies per parcel, then aggregates via guf_trust_inflow(). Schema maps directly to geo-data pipeline column names for zero-friction integration with GeoJSON/CSV sources. - collective.make_urban_collective(parcel_count) — Synthetic dense-urban archetype (75 % residential_primary · 15 % commercial_retail · 5 % commercial_office · 5 % institutional). - collective.make_rural_collective(parcel_count) — Synthetic rural archetype (50 % agricultural_active · 20 % agricultural_fallow · 20 % residential_primary · 10 % conservation). - calibration.py — Rate and weight calibration tools. - guf_rate_calibration(inventory, target_guf_levy_ratio, population, epsilon) — Closed-form linear solve to find the use-coefficient multiplier k such that aggregate GUF ≈ target × levy_revenue. Binary-verifies with a sample run; reports converged flag. - guf_lvi_weight_sensitivity(inventory, epsilon, weight_variants) — Sweeps Location Value Index weight configurations to quantify how sensitive aggregate GUF is to sub-index weighting choices (centrality vs. transit vs. services vs. natural amenity). Ships five canonical variants; callers may supply their own. - land/__init__.py re-exports all four new public functions.

Multi-period automation→levy→GUF stress scenario - scenarios/guf_stress.automation_levy_guf_stress(parcel_inventory, epsilon_start, epsilon_end, n_periods, ...) — Period-by-period fiscal stress loop: as ε rises, levy revenue (derived from the EOH pipeline) falls, GUF tracks the Ψ(ε) bell curve, and the sufficiency guarantee cost evolves. Reports levy_peak_period, guf_peak_period, crossover_period (first period where GUF exceeds levy), first_insolvency, compensation_adequacy, and outcome ADEQUATE / PARTIAL / CRISIS.

Multi-period simulation scenarios (hours_eoh/scenarios/) - long_run.py — canonical_arc_trajectory, trust_depletion_stress, automation_transition_trajectory: three functions that call run_simulation() and return multi-period trajectories with inflection-point detection and stressor profiles. - indust_overshoot.py — indust_overshoot_baseline, indust_recovery_trajectory: models the industrial-overshoot archetype (high capital age, degraded ecosystem, large deferred-ecological backlog) and recovery pathways. - shocks.py additions — labor_income_shock (income-fraction shock with solvency delta) and compound_shock (combines ecological, demographic, and automation shocks; combined outcome is always ≥ worst individual outcome).

Simulation engine extensions (hours_eoh/core/simulation.py) - simulate_period(..., workforce_epsilon_decay: bool = False) — Optional parameter; when True, workforce_fraction shrinks proportionally with rising ε. Default False preserves backward compatibility. - simulate_period(..., guf_net_inflow: float | None = None) — Optional GUF land-fee revenue injected into the Trust each period. - run_simulation() and the eoh simulate CLI forward both new parameters.

GUF calibration fix - GUF_USE_* constants in data.py multiplied by 100 (e.g. residential_primary 0.10 → 10.0 TEH/SLU/yr, commercial_retail 0.30 → 30.0). At ε = 0.40 with a 1 M-population territory (~420 k parcels), aggregate GUF is now co-equal with levy revenue — the design target in the mission statement. - guf_fiscal_integration() labor-income proxy replaced: was trust_balance × 0.5 (35× inflated); now uses eoh_to_teh_pipeline(epsilon, population)["teh_created"].

CLI (utils/) - guf inventory calculate --parcels FILE [--epsilon ε] [--median-income TEH] — Batch GUF from a JSON parcel file; prints aggregate summary table or JSON. - guf inventory sweep --parcels FILE [--epsilon-start ε] [--epsilon-end ε] [--steps N] — Sweeps aggregate GUF across the ε arc. - guf inventory stress --parcels FILE [--epsilon-start ε] [--epsilon-end ε] [--periods N] — Multi-period automation→GUF stress with per-period trajectory table. - scenario command: 10 new scenarios wired — canonical-arc, trust-depletion, automation-transition, indust-baseline, indust-recovery, labor-shock, compound-shock, guf-integration, guf-writedown, guf-sweep. - simulate command: --workforce-decay and --guf-inflow TEH flags.

Documentation and diagrams - MkDocs GitHub Pages site deployed at https://hours-source.github.io/HOURS/. - Five GUF Mermaid diagrams: calculation flow, LVI component weights, Ψ(ε) epsilon arc, Trust fund flow, and ecological write-down pathways. Rendered SVGs in docs/images/.

Fixed

  • eco-collapse-1 (closed) — Ecological collapse is now handled via the GUF layer (land/guf.py §9) rather than direct TEH destruction. Two pathways: restoration (V_s baselines reset to recovery target, revenue maintained) and abandonment (rebuilding surcharge R_b amortised over 50 years). Preventive monitoring via eoh_accumulation_warning() (§9.8). research/writedown.py re-exports with rationale.
  • guf_net_inflow guard in simulate_period() corrected from > 0.0 to is not None so negative inflows (subsidy-heavy periods) are no longer silently dropped.
  • WORKFORCE_FRACTION_MIN named constant added to data.py; replaces anonymous 0.05 literal in simulation.

Changed

  • Test count: 1040 → 1169 (129 new tests across tests/land/, tests/scenarios/).

0.1.0 — 2026-05-06

Initial public release of the HOURS EOH framework.

Added

Core package (hours_eoh/core/) - trajectory.py — Canonical arc, ε derivation, canonical_physical_state(), compute_epsilon() - eoh_generation.py — Four EOH domain functions: personal_eoh, infrastructure_eoh, ecological_eoh, knowledge_eoh, total_eoh - registration.py — Per-domain sigmoid admission curves; personal_eoh_registration_share (near-zero at ε=0), total_registration_share (labor composite) - eoh_fulfillment.py — EOH → TEH pipeline: human_eoh_share, human_eoh_per_domain, registered_eoh, eoh_to_teh_pipeline - multipliers.py — Condition II multiplier band and tier logic - fiscal.py — Fiscal architecture: levies, stewardship and ecological allocations (co-equal), sufficiency guarantee, trust management, care stipend - prices.py — Price dynamics tied to human labor content: basket_price, purchasing_power, floor_purchasing_power - capital.py — Asset and human capital lifecycle: make_asset, birth_event, maturation_update, death_event, writedown_trigger, execute_writedown - eoh_dynamics.py — Time-evolution: eoh_compounding, asset_condition_trajectory, regenerative_offset, eoh_reduction_ratio - population.py — Population structure, age distribution, demographic events (aging) - workforce.py — Workforce lifecycle, domain headcount, competency reserve, apply_death_redistribution - conditions.py — Structural Conditions I–IV enforcement: balance_check, condition_iii_balance_growth_check - dashboard.py — Condition monitors and EOH/fiscal health indicators - civilization.py — Endogenous ε from capital stock: civilization_epsilon, CAPITAL_MACHINE_PROFILES - simulation.py — Period simulation engine: make_economy_state, simulate_period

Land module (hours_eoh/land/) - guf.py — Ground Use Fee framework (NLSA TM-0042, 7th Ed.): 14 functions from epsilon_scaling through guf_trust_inflow

Scenarios (hours_eoh/scenarios/) - sweep.py — epsilon_sweep: arc coherence check with fiscal solvency at every ε - shocks.py — automation_failure_shock, demographic_shock, ecological_eoh_spike - maintenance.py — deferred_maintenance_crisis, care_registration_delay - recovery.py — maintenance_recovery_schedule, minimum_fulfillment_for_recovery - sensitivity.py — fiscal_parameter_sweep, eoh_arc_sensitivity, epsilon_delta_sensitivity

Research (hours_eoh/research/) - investment.py — rank_investment_candidates, optimal_investment, eoh_reduction_ratio (re-exported from core) - writedown.py — Placeholder: ecological write-down for collapse scenarios (eco-collapse-1, future work)

Test suite - 1040 tests across 20 test files covering the full arc from ε = 0 to ε = 0.99 - Phase tests (1–14) cover the complete development arc - Module tests cover trajectory, civilization ε, GUF, and all scenario modules

Design invariants established

  • ε is a derived observable, not a policy lever or generation-function input
  • EOH generation takes physical state; EOH fulfillment takes ε
  • data.py is the single source of truth for all named constants
  • Per-domain registration split: personal uses demand sigmoid; non-personal use labor composite
  • Ecological and stewardship allocations are co-equal (neither is residual)
  • Zero interest (Condition III): balances grow only through labor
  • All functions verified at ε = 0 (subsistence) and ε = 0.99 (effective post-scarcity)

Known gaps

  • eco-collapse-1 — Ecological write-down for collapse scenarios not yet implemented (placeholder in research/writedown.py)