leaf: q2-earth-atmospheric-ceiling pass: 01-research started: 2026-06-09T02:30:00Z ended: 2026-06-09T02:50:00Z researcher: claude-opus-4-7

Pass 01 — Research

Source gathering for q2 (Earth atmospheric ceiling). Five tier-S/A primary sources retrieved fresh; emission factors and current-cadence-impact baselines established.

Tier-grouped source list

Tier S — peer-reviewed primary

Tier A — peer-reviewed reviews / credentialed preprints

NOT YET FETCHED (flagged for re-pass)

• NASA TM-20240013276 Impact of Spaceflight on Earth's Atmosphere — PDF fetch returned binary streams; needs HTML alternative or alternative NTRS doc.
• Barker et al. 2026 Earth's Future on megaconstellation radiative forcing — paywalled (402).
• Direct fetch of Maloney 2025 JGR (only NOAA CSL secondary coverage retrieved).

Topics surfaced (add to topics.yaml at report level)

• rocket-ozone-depletion — direct ozone destruction from chlorine (SRM), black carbon (kerosene), NOx (reentry)
• stratospheric-water-vapor — H2O from methalox/LH2 combustion at stratospheric altitudes; HOx-cycle ozone effects
• reentry-alumina-mesosphere — Al2O3 from spacecraft ablation 40-70 km; polar mesospheric heating; polar vortex effects
• propellant-specific-emissions — per-kg emission factors by propellant type (Ryan 2022 canonical)
• polar-stratospheric-clouds — alumina/metals nucleation effects on PSCs (Murphy 2023, Maloney 2025)

Methalox-specific gap

The peer-reviewed literature lacks dedicated atmospheric modeling for a pure-methalox future. Revell 2025 explicitly notes methalox as a research gap. Ryan 2022's emission-factor table omits methalox (we can derive it from stoichiometry: 0.6 kg CO2 + 0.49 kg H2O + ~10 g NOx + minimal BC per kg propellant). At Starship-dominated cadence the ozone and aerosol picture should differ substantially from the Revell-2025 mixed-fuel projection — no chlorine, dramatically less BC, no alumina from combustion (alumina from reentry of payloads remains, however).

This is a real research gap. q2's calc pass must derive the methalox-specific atmospheric ceiling first-principles.

Coverage assessment

• Tier S is good: 4 primary peer-reviewed papers covering the four main atmospheric impact pathways (direct ozone, stratospheric H2O, reentry alumina, propellant emission factors).
• Tier A coverage thin: Kukreja 2025 covers the LCA frame but not specific cadence-thresholds. Could add IPCC AR6 Chapter 8 (atmospheric composition) for general radiative-forcing context.
• Tier B (public figures): Casey Handmer, Eli Dourado, Eloise Marais (academic atmosphere specialist) all have public positions. None directly compiled here; could be a re-pass target.
• Tier C/D: Industry trade coverage (Space.com, Phys.org) consolidates the same primary findings — not separately reviewed in the current pass.

Sub-pass deliverables

• 5 extract.md files in sources/ (all tier S/A)
• 1 pass-01-research.md (this file)
• 0 entries in claims.yaml — claims come from calc + reconcile

Anti-pattern hygiene check

• No claim-bearing prose in this pass ✓
• No source borrowing across passes yet — calc-pass will be sources-sealed ✓
• All extracts have proper frontmatter (tier, type, peer_reviewed, venue, year, topics) ✓
• No hallucinated sources — every extract corresponds to a fetched WebFetch or PMC retrieval ✓

Next sub-pass

Calc — derive Earth atmospheric ceiling first-principles, sources sealed. Apply lessons from q1: distinguish fundamental (atmospheric absorption capacity, ozone destruction rate per unit injection, irreversible thresholds) from willingness-to-pollute (current-cadence projections that scale linearly until they don't).

leaf: q2-earth-atmospheric-ceiling pass: 02-calc started: 2026-06-09T02:55:00Z ended: 2026-06-09T03:30:00Z researcher: claude-opus-4-7 sources_sealed: true

Pass 02 — Calc (sources sealed)

Methodology framing: per-launch emissions and per-pollutant atmospheric perturbation are derived from physics + vehicle spec (stoichiometry, atmospheric mass-budget arithmetic, ozone catalytic-cycle efficiencies). Atmospheric "ceilings" are stated in three tiers: detectable (5% perturbation; current scientific literature begins to flag effects), disruptive (50% perturbation; large-scale ecosystem and climate consequences), catastrophic (>100% perturbation; stratosphere is essentially anthropogenically dominated, baseline atmospheric chemistry replaced). The pass is sources sealed — no reading of sources/ during this analysis.

Per the q1 lessons:

• Distinguish fundamental from willingness. Atmospheric chemistry is fundamentally physical: pollutants accumulate and degrade absorption capacity regardless of capital/willingness.
• Use proper units consistently. Tt = teratonnes = 10¹² t = 10¹⁵ kg; Gt = gigatonnes = 10⁹ t = 10¹² kg.
• Don't extrapolate naively from current data; first-principles the atmospheric absorption capacity.

Assumptions

1. Reference vehicle: Block 3 Starship + Super Heavy. 5,650 t total propellant; LOX:CH4 = 3.6:1 mass; 100 t payload to LEO reusable.
2. Methalox combustion stoichiometry: CH4 + 2 O2 → CO2 + 2 H2O. Mass balance: 16 + 64 = 44 + 36. Per kg CH4: 2.75 kg CO2 + 2.25 kg H2O.
3. Mixture-ratio fuel-rich: at 3.6:1 (vs 4:1 stoichiometric), ~10% of CH4 doesn't fully combust to CO2 — some becomes CO, some unburned. Treat as ~5% CO2 deficit, ~5% unburned CH4 (a potent GHG); the H2O figure is robust to this since both H from CH4 and from incomplete combustion goes to H2O.
4. NOx emission factor: 10-50 g per kg propellant for non-SRM engines (atmospheric N2 dissociation in combustion). Use 20 g/kg as central estimate for methalox.
5. BC emission factor: ~0.1-1 g per kg propellant for methalox (much less than kerosene's 35 g/kg per literature). Use 0.5 g/kg.
6. Stratospheric injection fraction: ~50% of propellant combustion products end up above ~12 km (stratosphere/mesosphere). Below 12 km mixes into troposphere with ~10-day residence (short-lived effects). Stratospheric residence ~5 years (long-lived effects).
7. Atmospheric reservoir masses (current, baseline):
   • Total atmosphere: 5.15 × 10¹⁸ kg = 5,150 Tt
   • Stratospheric mass (~10-50 km, by pressure-weighted mass): ~17% of total ≈ 880 Tt (more conservative: pressure-weighted column above tropopause)
   • Stratospheric H2O (current, ~3 ppm mass): ~2.6 Gt
   • Stratospheric ozone (total column ~3 mm STP × Earth area): ~3 Gt
   • Stratospheric NOx (steady state): ~0.5 Mt
8. Reentry NOx production (CORRECTED post-audit): Larson 2017 / Vliex et al. 2024 / Ryan 2022 converge on 17.5% of reentered mass as NOx for reusable spacecraft (40% for fully discarded objects undergoing complete burn-up). Use 0.175 kg NOx per kg reentered mass. Original calc draft used 5 kg/kg — corrected by factor 30 per Codex audit.
9. Reusable architecture: Block 3 vehicle returns; only payload + minor expendable mass reenters per launch. Per launch reentered mass ≈ 100 t (payload + booster recovery losses). Reusable vehicle adds ~5 t/launch ablation/wear.
10. Pollutant residence times in stratosphere: ~5 years for H2O and aerosols; NOx is much shorter-lived (days-to-weeks in lower stratosphere, months in middle/upper) — box-model with 5-year residence is dimensionally wrong for NOx and overstates accumulation. Use box-residence only for H2O / aerosol species. NOx perturbation should be evaluated against instantaneous catalytic-cycle ozone-destruction rates, not steady-state mass accumulation.

Step 1 — per-launch emission products (Block 3 Starship, methalox)

# Block 3 Starship per-launch propellant breakdown
propellant_total_t = 5650
lch4_t = propellant_total_t / 4.6      # 1228 t methane
lox_t = propellant_total_t * 3.6/4.6   # 4422 t oxygen

# Methalox stoichiometry: CH4 + 2O2 -> CO2 + 2H2O
# Per kg CH4 (fully combusted): 2.75 kg CO2 + 2.25 kg H2O
# At 3.6:1 mixture ratio (sub-stoichiometric), ~5% deficit on full combustion
combustion_efficiency = 0.95
co2_per_launch_t = lch4_t * 2.75 * combustion_efficiency      # 3208 t CO2
h2o_per_launch_t = lch4_t * 2.25                              # 2763 t H2O (H all converts even sub-stoich)
unburned_ch4_t = lch4_t * (1 - combustion_efficiency)         # 61 t CH4 unburned

# Per-kg-propellant emission factors
nox_g_per_kg_prop = 20    # methalox central estimate
bc_g_per_kg_prop = 0.5    # methalox is very clean burning
nox_per_launch_t = propellant_total_t * nox_g_per_kg_prop / 1000  # 113 t NOx
bc_per_launch_t = propellant_total_t * bc_g_per_kg_prop / 1000     # 2.8 t BC

# Stratospheric fraction (above ~12 km tropopause)
strat_frac = 0.50
strat_co2_per_launch_t = co2_per_launch_t * strat_frac           # 1604 t
strat_h2o_per_launch_t = h2o_per_launch_t * strat_frac           # 1381 t
strat_nox_per_launch_t = nox_per_launch_t * strat_frac           # 57 t
strat_bc_per_launch_t = bc_per_launch_t * strat_frac             # 1.4 t

# Reentry NOx (CORRECTED): 17.5% of reentered mass per Larson 2017 / Vliex 2024 / Ryan 2022
# Reusable Block-3 Starship: ~100 t payload + ~5 t booster recovery loss per launch
reentry_mass_t = 100
nox_per_reentry_t = reentry_mass_t * 0.175    # 17.5 t NOx per launch from reentry heat
# (was previously calculated as 500 t at 5 kg/kg; corrected by factor 30)

Step 2 — atmospheric reservoir baselines (steady-state)

strat_h2o_baseline_Gt = 2.6        # ~3 ppm mass × 880 Tt strat mass
strat_o3_baseline_Gt = 3.0         # total column × Earth area
strat_nox_baseline_Mt = 0.5
residence_yr = 5

Step 3 — perturbation per launch rate

def steady_state_addition_Mt(per_launch_t, launches_per_yr):
    """Steady-state stratospheric load from emission with 5-yr residence."""
    annual_addition_Mt = per_launch_t * launches_per_yr / 1e6
    steady_state_Mt = annual_addition_Mt * residence_yr
    return steady_state_Mt

scenarios = [
    ("250 (current 2023)",  250),
    ("2,000 (Revell ambit)", 2000),
    ("10,000 (1 Mt LEO/yr)", 10000),
    ("100,000 (10 Mt LEO/yr)", 100000),
    ("1,000,000 (100 Mt LEO/yr)", 1000000),
    ("10,000,000 (1 Gt LEO/yr)", 10000000),
    ("100,000,000 (10 Gt LEO/yr)", 100000000),
]

print(f'{"cadence":<28} | {"strat H2O ss":>14} | {"% baseline":>10} | {"strat NOx ss":>14} | {"% baseline":>10}')
for label, N in scenarios:
    h2o_ss = steady_state_addition_Mt(strat_h2o_per_launch_t, N)
    h2o_pct = 100 * h2o_ss / (strat_h2o_baseline_Gt * 1000)
    nox_ss = steady_state_addition_Mt(strat_nox_per_launch_t + nox_per_reentry_t, N)
    nox_pct = 100 * nox_ss / strat_nox_baseline_Mt
    print(f'{label:<28} | {h2o_ss:>10.1f} Mt | {h2o_pct:>9.2f}% | {nox_ss:>10.1f} Mt | {nox_pct:>9.0f}%')

Step 4 — verify with python

python3 -c "
lch4_t = 5650/4.6
co2 = lch4_t * 2.75 * 0.95
h2o = lch4_t * 2.25
nox = 5650 * 0.020
bc = 5650 * 0.0005
strat_frac = 0.50
h2o_strat = h2o * strat_frac
nox_strat = nox * strat_frac
nox_reentry = 100 * 5
print(f'CH4: {lch4_t:.0f} t   |  CO2: {co2:.0f} t  |  H2O: {h2o:.0f} t  |  NOx: {nox:.0f} t  |  BC: {bc:.1f} t')
print(f'Strat H2O: {h2o_strat:.0f} t  |  Strat NOx: {nox_strat:.0f} t  |  Reentry NOx: {nox_reentry:.0f} t')
print()
strat_h2o_baseline_Gt = 2.6
strat_nox_baseline_Mt = 0.5
residence = 5
print(f'{\"cadence\":<28} | {\"strat H2O ss (Mt)\":>18} | {\"% baseline\":>10} | {\"strat NOx ss (Mt)\":>18} | {\"% baseline\":>10}')
for label, N in [
    ('250 (current 2023)', 250),
    ('2,000 (Revell ambit)', 2000),
    ('10,000 (1 Mt LEO/yr)', 10000),
    ('100,000 (10 Mt LEO/yr)', 100000),
    ('1,000,000 (100 Mt LEO/yr)', 1000000),
    ('10,000,000 (1 Gt LEO/yr)', 10000000),
    ('100,000,000 (10 Gt LEO/yr)', 100000000),
]:
    h2o_ss = h2o_strat * N / 1e6 * residence
    h2o_pct = 100 * h2o_ss / (strat_h2o_baseline_Gt * 1000)
    nox_total = nox_strat + nox_reentry
    nox_ss = nox_total * N / 1e6 * residence
    nox_pct = 100 * nox_ss / strat_nox_baseline_Mt
    print(f'{label:<28} | {h2o_ss:>14.2f} Mt | {h2o_pct:>9.3f}% | {nox_ss:>14.2f} Mt | {nox_pct:>9.1f}%')
"

Run output:

CH4: 1228 t   |  CO2: 3208 t  |  H2O: 2763 t  |  NOx: 113 t  |  BC: 2.8 t
Strat H2O: 1382 t  |  Strat NOx: 57 t  |  Reentry NOx: 500 t

cadence                      | strat H2O ss (Mt)  | % baseline | strat NOx ss (Mt)  | % baseline
250 (current 2023)           |           1.73 Mt |     0.066% |           0.70 Mt |     139.3%
2,000 (Revell ambit)         |          13.82 Mt |     0.532% |           5.57 Mt |    1114.0%
10,000 (1 Mt LEO/yr)         |          69.10 Mt |     2.658% |          27.85 Mt |    5570.0%
100,000 (10 Mt LEO/yr)       |         691.00 Mt |    26.577% |         278.50 Mt |   55700.0%
1,000,000 (100 Mt LEO/yr)    |        6910.00 Mt |   265.77 % |        2785.00 Mt |  557000.0%
10,000,000 (1 Gt LEO/yr)     |       69100.0  Mt |  2657.7 %  |       27850.0  Mt | 5570000.0%
100,000,000 (10 Gt LEO/yr)   |      691000.0  Mt | 26577.0 %  |      278500.0  Mt | 55700000.0%

Step 5 — anchor to peer-reviewed cadence-scenarios (post-audit revision)

Codex audit flagged that linear extrapolation from Revell 2025's 2,040-launches/yr (mixed-fuel) scenario to my "30-50% ozone loss" at 10⁵-10⁶ launches/yr is unsupported. The actually-modeled high-cadence scenario in the literature is Larson 2017 (high-rate reusable hydrogen-propellant flights), which is the closest analog to a future methalox-dominated regime (both lack chlorine and produce dominant H2O + NOx perturbations).

Larson 2017 anchored cadence → ozone-loss results:

The original step-5 perturbation table that I produced (linear scaling Revell 2025 → -30% global ozone at 10⁵ launches/yr) overstated the loss because NOx and HOx ozone-destruction cycles saturate at high pollutant concentrations. Larson 2017's direct chemistry-climate modeling of high-cadence hydrogen launches finds much lower destruction per launch than my linear scaling implied. The ozone-saturation effect is itself part of why the ceiling moves up by ~1-2 OOM from my original estimate.

Step 5 (legacy, kept for traceability) — interpret perturbations at each cadence

Note on NOx percentages: the very large percentages (>100% perturbation) reflect that reentry NOx alone at modest cadence saturates stratospheric NOx. Reentry NOx is the dominant ozone-loss mechanism, not propellant emissions — and it's largely independent of propellant choice. NOx-catalyzed ozone destruction is saturable: at very high NOx the catalytic cycles approach a different chemistry regime, but at the >1000% perturbation level you have lost most ozone via the standard cycles.

Three-tier atmospheric ceiling

Defining tiers by stratospheric perturbation level:

• Detectable (5% perturbation): atmospheric chemistry literature begins to flag concrete effects.
• Disruptive (50% perturbation): large biosphere and climate consequences (substantial ozone loss, ecosystem stress).
• Catastrophic (>100% perturbation): stratosphere fundamentally anthropogenic; baseline chemistry replaced.

The NOx ceiling is strikingly low: reentry NOx at even 100 launches/yr already meaningfully perturbs stratospheric NOx, because the natural stratospheric NOx baseline is small (~0.5 Mt) and per-launch reentry NOx is substantial (~500 t).

But NOx perturbation doesn't directly equal ozone loss percentage — Ryan 2022 finds reentry NOx accounts for ~51% of rocket-driven ozone loss at current rates. The ozone-catalysis efficiency of stratospheric NOx is bounded by HOx, ClOx, and BrOx couplings; saturating NOx doesn't 1:1 saturate ozone destruction.

A more conservative atmospheric ceiling, defined by operationally significant ozone loss:

• 5% global ozone loss (broadly catastrophic for ecosystems): roughly 100× Revell 2025 baseline (-0.29% at 2,040/yr) → ~200,000 launches/yr (~20 Mt LEO/yr).
• 50% global ozone loss (UV sterilization at the surface in tropics): roughly 1000× Revell → ~2,000,000 launches/yr (~200 Mt LEO/yr).

These rates assume methalox-dominant fuel mix (eliminating Cl from SRMs; greatly reduced BC). Under the current mixed-fuel projection (Revell 2025), the ceilings are ~10-50× lower because Cl and BC dominate at low cadences.

Result — q2 atmospheric ceiling (post-audit revised)

Earth chemical-rocket throughput ceiling from atmospheric chemistry, methalox-dominant regime, anchored on Larson 2017:

• Detectable global effects (~0.5% ozone loss): ~10⁵ launches/yr (~10 Mt/yr LEO)
• Disruptive (~3-5% ozone loss, substantial Antarctic loss, biosphere UV stress): ~10⁶ launches/yr (~100 Mt/yr LEO)
• Catastrophic (~10-30% global ozone loss, UV catastrophe at high latitudes): ~10⁷ launches/yr (~1 Gt/yr LEO)
• Beyond ceiling (stratospheric chemistry fundamentally altered): ~10⁸ launches/yr (~10 Gt/yr LEO)

The atmospheric ceiling sits at ~10⁶-10⁷ launches/yr (100 Mt/yr to 1 Gt/yr LEO). Compared to the q1 solar-PV ceiling at ~100 Gt/yr LEO: q2 binds 10-100× lower than q1. Atmospheric chemistry remains the load-bearing Earth-side constraint, though by a smaller margin than my original Mt/yr ceiling claimed (corrected from ~1,000-10,000× to ~10-100× lower than q1).

At cosmic Tt/yr (10¹² t/yr = 10¹⁰ launches/yr) LEO scale, atmospheric chemistry binds by 3-4 orders of magnitude — completely impossible from Earth chemical alone. The Moon's architectural necessity at cosmic scale is established by q2.

At Mt-Gt/yr scale, the Moon's value rests primarily on q2 atmospheric grounds + q1 solar-PV grounds jointly; q2 binds first (at ~100 Mt/yr) while q1 binds at ~10-100 Gt/yr.

Confidence

• High confidence in the per-launch methalox emission factors (stoichiometry + Ryan 2022 NOx factor).
• Medium confidence in the reentry NOx emission factor (5 kg/kg-reentered is rough; literature range 1-10 kg/kg).
• High confidence in the qualitative ordering: NOx and H2O bind well before alumina; methalox H2O is a real ozone-relevant perturbation at high cadence.
• Medium-high confidence in the specific cadence-thresholds (~10⁴-10⁶ launches/yr range for "disruptive"). Exact threshold depends on what level of biosphere damage we accept.
• High confidence that the q2 atmospheric ceiling is the load-bearing Earth-side constraint at any Mt/yr-scale LEO mass throughput.

Caveats

• The Maloney 2025 reentry alumina projection assumes aluminum-skinned satellites; if megaconstellation hardware moves to other materials, alumina injection decreases.
• The reusable-Starship architecture reduces vehicle reentry mass per launch; the dominant reentry mass comes from payloads (satellites). Reentry NOx scales with total reentry mass, which is partly architecture-dependent.
• Polar geometry: ozone effects concentrate at high latitudes (polar vortex chemistry). Equatorial ozone may be more resilient than the global-average figures suggest, but high-latitude effects extend to mid-latitudes via atmospheric circulation.
• Synthetic-methane combustion is chemically identical to fossil-methane combustion. The Handmer solar-abundance regime does NOT reduce per-launch atmospheric chemistry — it just changes where the carbon came from.
• Long-term residence times in the stratosphere are ~5 years for H2O and ~years for aerosols. Reaching steady-state at a new launch rate takes ~5-10 years; transient effects are smaller.

Implication for the report

q2 establishes the Earth-side throughput ceiling at ~Mt-Gt/yr LEO scale on atmospheric-chemistry grounds. This is the binding constraint on Earth-launched-chemical-rocket throughput at any post-current cadence — sitting 1,000-10,000× below q1's solar-PV ceiling.

The Moon's architectural necessity at cosmic Tt/yr scale is overwhelmingly established by q2. It's also established at the much more modest 10 Mt-Gt/yr scale where serious solar-system industrialization wants to operate. The q2 atmospheric ceiling is therefore the actually-load-bearing Earth-side answer.

q1's solar-PV-area ceiling matters only as a floor below which the atmospheric ceiling doesn't matter. At any cadence above ~10⁴ launches/yr, atmospheric chemistry binds. q9 synthesis should lean on q2, not q1.

Derived claims (to add to claims.yaml in pass-03-reconcile)

• q2.c1 [derived, high]: Methalox per-launch stratospheric injection: ~1,400 t H2O + 1,600 t CO2 + 60 t NOx + 3 t BC + 500 t reentry-NOx. At 250 launches/yr (current 2023 cadence), stratospheric H2O perturbation 0.07%, NOx +140% (reentry-dominated).
• q2.c2 [derived, high]: Atmospheric chemistry ceiling for methalox Earth-launch: detectable at ~10⁴ launches/yr (~1 Mt/yr LEO), disruptive at ~10⁵ launches/yr (~10 Mt/yr LEO), catastrophic at ~10⁶-10⁷ launches/yr (~100 Mt - 1 Gt/yr LEO).
• q2.c3 [derived, high]: Reentry NOx (not propellant NOx) is the dominant ozone-loss driver in methalox-dominant futures; reentry NOx is propellant-independent because it comes from atmospheric N₂ dissociation during reentry heating.
• q2.c4 [derived, high]: Synthetic methane via Sabatier in the Handmer-solar-abundance regime does NOT reduce per-launch atmospheric chemistry. Combustion products are chemically identical to fossil-CH4 combustion. The propellant-source change improves carbon-cycle but doesn't help q2.
• q2.c5 [derived, medium-high]: At cosmic Tt/yr LEO (10¹² t/yr = 10¹⁰ launches/yr), atmospheric metrics are perturbed by 10⁴-10⁸× — physically impossible. Earth chemical-rocket cannot operate at cosmic Tt/yr because atmosphere itself is being fundamentally rebuilt.
• q2.c6 [derived, high]: The q2 atmospheric ceiling (~10⁵-10⁶ launches/yr at "disruptive" tier) is 1,000-10,000× lower than the q1 solar-PV ceiling (~10⁹ launches/yr). Atmospheric chemistry is the load-bearing Earth-side constraint at any Mt/yr-scale LEO mass throughput.

Next sub-pass

--sub reconcile — unseal sources, compare derived numbers to Revell 2025, Murphy 2023, Maloney 2025, Ryan 2022, Kukreja 2025.

leaf: q2-earth-atmospheric-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Response to Codex audit on q2 pass-02-calc

Codex caught four high-severity issues plus four medium. Most are accepted. The qualitative conclusion (atmospheric chemistry binds well before solar-PV area) holds, but the quantitative ceiling moves up by ~1-2 orders of magnitude.

Accepted — substantive corrections

• Reentry NOx 5 kg/kg-reentered (HIGH, contradicted): Codex cites Larson 2017, Vliex et al. 2024, Ryan 2022 all using ~17.5% of reentered mass as NOx (0.175 kg/kg) for reusable spacecraft, with 40% for discarded objects. My 5 kg/kg was ~30× too high. Per Block-3 Starship launch at 100 t reentered: 17.5 t NOx (not 500 t).

• Catastrophic ozone scaling (HIGH, contradicted): Larson 2017 modeled 10⁵-10⁶ launches/yr with hydrogen propellant and found ~0.5% global ozone loss at 10⁵ and ~11 DU (~3-4%) at 10⁶ launches/yr — not the "30-50% ozone loss" I extrapolated. NOx and HOx-cycle catalytic destruction saturates at high concentrations rather than scaling linearly with emissions. My linear extrapolation of Revell 2025 was wrong.

• Revell 2025 extrapolation (HIGH, unsupported): Revell's depletion is driven by Cl (from SRMs) and BC (from kerosene), NOT by NOx. Revell explicitly notes rocket-emitted NOx is insignificant in their setup AND they omit reentry NOx. Linear scaling from Revell to a methalox-only future at 10⁵-10⁶ launches/yr is unjustified.

• NOx perturbation table internal inconsistency (HIGH): my table said "5% NOx perturbation at ~90 launches/yr" but if 250 launches → 140%, then 5% binds at ~9 launches/yr. Math error in my output.

• NOx 20 g/kg propellant (MEDIUM): newer altitude-resolved inventories show NOx EI is altitude-dependent: 33 g/kg at ground falling to <1 g/kg above 14 km. The bulk-stratospheric methalox NOx contribution is much smaller than I claimed.

• NOx residence-time model (MEDIUM): NOx is short-lived in the lower stratosphere (days-weeks), not 5 years. My 5-year box-model accumulation is dimensionally wrong for NOx.

• Stratospheric injection fraction (MEDIUM): Vliex et al. 2024 splits propellant burn ~20-29% in stratosphere, ~10-19% in mesosphere, balance higher up — not the 50% I assumed.

• Combustion model (MEDIUM): fuel-rich methalox produces CO + H₂ intermediates with afterburning, not "5% CO2 deficit + 5% unburned CH4." H₂O figure stays close; CO2 figure may be lower.

Recomputed q2 ceiling

Anchoring on Larson 2017 (high-cadence H2-propellant scenarios, the closest analog to a methalox-dominant future since both lack chlorine and produce dominant H2O + NOx perturbations):

Recomputed atmospheric ceiling: ~10⁶-10⁷ launches/yr (100 Mt/yr to 1 Gt/yr LEO) — about 1-2 OOM higher than my original Mt/yr estimate, but still 10-100× lower than q1's solar-PV ceiling at ~100 Gt/yr LEO.

The qualitative conclusion holds: atmospheric chemistry is the load-bearing Earth-side constraint at any post-current cadence. Cosmic Tt/yr (10¹² t/yr = 10¹⁰ launches/yr) is still unreachable from Earth chemical — atmospheric chemistry binds by 3-4 OOM at cosmic scale.

Reentry alumina ceiling (Murphy 2023 / Maloney 2025) stays at ~10⁵-10⁶ launches/yr — but largely driven by aluminum-skinned payloads, not vehicle propellant. With non-Al satellite materials and reusable vehicles, this ceiling is partly architecture-dependent.

Disputed / clarified

• Methalox vs fossil/synthetic methane (MED, partial): my original claim was specifically about synthetic-vs-fossil methane combustion being chemically identical. Codex right that it could be misread as methalox-vs-current-fuel-mix (which is dramatically different — methalox eliminates Cl, alumina, and most BC). CLARIFY: synthetic-vs-fossil-methane is identical combustion; methalox-vs-kerosene-SRM is much cleaner. Both true; readers need both.

What changes in claims.yaml

• q2.c1 — per-launch emissions: reentry NOx corrected from 500 t → 17.5 t (factor 30 lower). H2O figure ~1,400 t stays. NOx propellant figure adjusted with altitude dependence.
• q2.c2 — atmospheric ceiling: shift "disruptive" from 10⁵ launches/yr to 10⁶, "catastrophic" from 10⁶-10⁷ to 10⁷-10⁸.
• q2.c3 — reentry NOx dominant: still load-bearing but with corrected magnitude. At Larson 17.5%, reentry NOx ≈ propellant NOx for methalox at typical reentry mass.
• q2.c5 — Tt/yr still unreachable: holds with corrected numbers. Tt/yr cadence = 10¹⁰ launches/yr, ~3-4 OOM above corrected catastrophic ceiling.
• q2.c6 — atmospheric ceiling vs PV ceiling: q2 still 10-100× lower than q1 (not 1,000-10,000× as I originally claimed). Atmospheric chemistry remains the binding constraint, but by a smaller margin.

What does NOT change qualitatively

• Earth chemical-rocket throughput is atmospheric-chemistry-bound below the solar-PV-area ceiling.
• Cosmic Tt/yr LEO is unreachable from Earth chemical, primarily on atmospheric grounds (with PV-area as secondary constraint).
• Methalox is dramatically lower per-launch atmospheric impact than the current mixed fuel mix (no Cl from SRMs, much less BC, no alumina from combustion).
• Reentry NOx and reentry alumina are propellant-independent — they come from atmospheric N₂ dissociation and spacecraft material ablation during reentry heating.
• The Moon's architectural necessity for cosmic-scale throughput is established by q2 (atmospheric) + q1 (solar-PV) jointly.

Next action

Patch pass-02-calc.md with the corrected reentry NOx factor + Larson 2017 anchor + revised ceiling. Then reconcile (now needs Larson 2017 fetched as a source).

overall:
  verdict: weak
  summary: "The stoichiometric H2O arithmetic is mostly usable, but the NOx-driven ceiling is not defensible. The reentry-NOx factor, NOx residence/reservoir treatment, and Revell-to-catastrophe extrapolation are the load-bearing failures."

findings:
  - target: "q2.c1 methalox stoichiometry"
    target_kind: claim
    verdict: partial
    reason: "CH4 + 2 O2 -> CO2 + 2 H2O is correct, and the H2O mass is close. But the 3.6:1 fuel-rich treatment is oversimplified: incomplete combustion would produce substantial CO/H2 before afterburning, not simply a 5% CO2 deficit plus 5% unburned CH4."
    severity: medium

  - target: "NOx 20 g/kg-propellant"
    target_kind: claim
    verdict: weak
    reason: "Older rocket inventories use launch NOx factors in this rough range, but newer gridded inventories apply altitude-dependent afterburning: indirect NOx falls from 33 g/kg at ground to under 1 g/kg above 14 km. Treating 20 g/kg as a bulk stratospheric methalox factor is not supported. Source: https://www.nature.com/articles/s41597-024-03910-z"
    severity: medium

  - target: "reentry NOx 5 kg/kg-reentered"
    target_kind: claim
    verdict: contradicted
    reason: "This is too high by over an order of magnitude for spacecraft-style reentry. Larson/Park give about 17.5% of spacecraft mass as NOx; Vliex et al. use 17.5% for reusable objects and 40% for discarded objects, while Ryan 2022 used 17.5% for reusable and 100% only for complete burn-up cases. Source: https://repository.library.noaa.gov/view/noaa/21694/noaa_21694_DS1.pdf and https://www.nature.com/articles/s41597-024-03910-z"
    severity: high

  - target: "stratospheric reservoir baselines"
    target_kind: section
    verdict: partial
    reason: "O3 ~3 Gt and stratospheric H2O order-of-magnitude are acceptable. The NOx baseline is not a sound reservoir for this calculation: NOx/NOy chemistry is short-lived, altitude/season dependent, and not well represented by a 0.5 Mt box with 5-year accumulation."
    severity: medium

  - target: "50% stratospheric injection fraction"
    target_kind: claim
    verdict: partial
    reason: "A 50% above-tropopause fraction is plausible as a crude H2O/CO2 screen, but vertical placement matters. Recent inventories put about 20-29% of propellant burn in the stratosphere, about 10-19% in the mesosphere, and a large fraction above 80 km depending on profile; it is not valid for NOx because the NOx emission index is altitude dependent. Source: https://www.nature.com/articles/s41597-024-03910-z"
    severity: medium

  - target: "perturbation ratios"
    target_kind: section
    verdict: partial
    reason: "The scenario-table arithmetic follows from the stated inputs, but the inputs are wrong for NOx. The later NOx threshold table is internally inconsistent by about 10x: if 250 launches gives +140%, then 5% occurs near 9 launches/yr, not 90."
    severity: high

  - target: "Revell 2025 extrapolation"
    target_kind: claim
    verdict: unsupported
    reason: "Revell's 2040-launch result is for a mixed-fuel future and is driven mainly by chlorine chemistry and black-carbon radiative/dynamical effects; Revell explicitly says rocket-emitted NOx is insignificant in that setup and omits reentry NOx. Linear scaling to -30% or 50-80% global ozone loss is not justified. Source: https://www.nature.com/articles/s41612-025-01098-6"
    severity: high

  - target: "catastrophic ozone levels at 1e5-1e6 launches/yr"
    target_kind: claim
    verdict: contradicted
    reason: "Larson 2017 modeled 1e5-1e6 high-rate reusable hydrogen flights and found about 0.5% global ozone loss at 1e5 and about 11 DU at 1e6, not tens of percent global ozone loss. Starship/methalox is not identical, but this directly undercuts the claimed ozone-catastrophe scaling. Source: https://repository.library.noaa.gov/view/noaa/21694"
    severity: high

  - target: "missed atmospheric pathways"
    target_kind: topic
    verdict: partial
    reason: "The pass mentions some missing pathways but does not integrate them: BC-driven stratospheric heating, HOx/NOx coupling, PSC changes from H2O, mesospheric H2O/cloud effects, alumina/metal oxides from reentry, and material-dependent ablation chemistry. These are model problems, not simple reservoir ratios."
    severity: medium

  - target: "methalox vs fossil/synthetic methane"
    target_kind: claim
    verdict: partial
    reason: "Correct if the comparison is fossil methane versus synthetic methane: the combustion plume chemistry is essentially the same. Incorrect if generalized to methalox versus fossil kerosene/SRMs: methalox avoids chlorine/alumina from SRMs and likely reduces BC relative to kerosene, though methane BC factors remain uncertain."
    severity: medium

notes:
  - issue: "The final atmospheric ceiling may still be below the q1 solar-PV ceiling on H2O/BC/reentry grounds, but this pass does not prove the stated 1e4-1e6 launch/yr ozone ceiling."
    severity: high
  - issue: "Replace the NOx box model with literature emission factors by object class and altitude, then compare against chemistry-climate model outputs rather than percent-of-reservoir thresholds."
    severity: high
  - issue: "Sources consulted include Revell 2025, Ryan 2022, Larson 2017, Vliex/Scientific Data 2024, and Maloney 2022: https://www.nature.com/articles/s41612-025-01098-6 ; https://discovery.ucl.ac.uk/id/eprint/10150421/ ; https://repository.library.noaa.gov/view/noaa/21694 ; https://www.nature.com/articles/s41597-024-03910-z ; https://repository.library.noaa.gov/view/noaa/53971"
    severity: low

leaf: q2-earth-atmospheric-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Response to Codex audit on q2 pass-02-calc

Codex caught four high-severity issues plus four medium. Most are accepted. The qualitative conclusion (atmospheric chemistry binds well before solar-PV area) holds, but the quantitative ceiling moves up by ~1-2 orders of magnitude.

Accepted — substantive corrections

• Reentry NOx 5 kg/kg-reentered (HIGH, contradicted): Codex cites Larson 2017, Vliex et al. 2024, Ryan 2022 all using ~17.5% of reentered mass as NOx (0.175 kg/kg) for reusable spacecraft, with 40% for discarded objects. My 5 kg/kg was ~30× too high. Per Block-3 Starship launch at 100 t reentered: 17.5 t NOx (not 500 t).

• Catastrophic ozone scaling (HIGH, contradicted): Larson 2017 modeled 10⁵-10⁶ launches/yr with hydrogen propellant and found ~0.5% global ozone loss at 10⁵ and ~11 DU (~3-4%) at 10⁶ launches/yr — not the "30-50% ozone loss" I extrapolated. NOx and HOx-cycle catalytic destruction saturates at high concentrations rather than scaling linearly with emissions. My linear extrapolation of Revell 2025 was wrong.

• Revell 2025 extrapolation (HIGH, unsupported): Revell's depletion is driven by Cl (from SRMs) and BC (from kerosene), NOT by NOx. Revell explicitly notes rocket-emitted NOx is insignificant in their setup AND they omit reentry NOx. Linear scaling from Revell to a methalox-only future at 10⁵-10⁶ launches/yr is unjustified.

• NOx perturbation table internal inconsistency (HIGH): my table said "5% NOx perturbation at ~90 launches/yr" but if 250 launches → 140%, then 5% binds at ~9 launches/yr. Math error in my output.

• NOx 20 g/kg propellant (MEDIUM): newer altitude-resolved inventories show NOx EI is altitude-dependent: 33 g/kg at ground falling to <1 g/kg above 14 km. The bulk-stratospheric methalox NOx contribution is much smaller than I claimed.

• NOx residence-time model (MEDIUM): NOx is short-lived in the lower stratosphere (days-weeks), not 5 years. My 5-year box-model accumulation is dimensionally wrong for NOx.

• Stratospheric injection fraction (MEDIUM): Vliex et al. 2024 splits propellant burn ~20-29% in stratosphere, ~10-19% in mesosphere, balance higher up — not the 50% I assumed.

• Combustion model (MEDIUM): fuel-rich methalox produces CO + H₂ intermediates with afterburning, not "5% CO2 deficit + 5% unburned CH4." H₂O figure stays close; CO2 figure may be lower.

Recomputed q2 ceiling

Anchoring on Larson 2017 (high-cadence H2-propellant scenarios, the closest analog to a methalox-dominant future since both lack chlorine and produce dominant H2O + NOx perturbations):

Recomputed atmospheric ceiling: ~10⁶-10⁷ launches/yr (100 Mt/yr to 1 Gt/yr LEO) — about 1-2 OOM higher than my original Mt/yr estimate, but still 10-100× lower than q1's solar-PV ceiling at ~100 Gt/yr LEO.

The qualitative conclusion holds: atmospheric chemistry is the load-bearing Earth-side constraint at any post-current cadence. Cosmic Tt/yr (10¹² t/yr = 10¹⁰ launches/yr) is still unreachable from Earth chemical — atmospheric chemistry binds by 3-4 OOM at cosmic scale.

Reentry alumina ceiling (Murphy 2023 / Maloney 2025) stays at ~10⁵-10⁶ launches/yr — but largely driven by aluminum-skinned payloads, not vehicle propellant. With non-Al satellite materials and reusable vehicles, this ceiling is partly architecture-dependent.

Disputed / clarified

• Methalox vs fossil/synthetic methane (MED, partial): my original claim was specifically about synthetic-vs-fossil methane combustion being chemically identical. Codex right that it could be misread as methalox-vs-current-fuel-mix (which is dramatically different — methalox eliminates Cl, alumina, and most BC). CLARIFY: synthetic-vs-fossil-methane is identical combustion; methalox-vs-kerosene-SRM is much cleaner. Both true; readers need both.

What changes in claims.yaml

• q2.c1 — per-launch emissions: reentry NOx corrected from 500 t → 17.5 t (factor 30 lower). H2O figure ~1,400 t stays. NOx propellant figure adjusted with altitude dependence.
• q2.c2 — atmospheric ceiling: shift "disruptive" from 10⁵ launches/yr to 10⁶, "catastrophic" from 10⁶-10⁷ to 10⁷-10⁸.
• q2.c3 — reentry NOx dominant: still load-bearing but with corrected magnitude. At Larson 17.5%, reentry NOx ≈ propellant NOx for methalox at typical reentry mass.
• q2.c5 — Tt/yr still unreachable: holds with corrected numbers. Tt/yr cadence = 10¹⁰ launches/yr, ~3-4 OOM above corrected catastrophic ceiling.
• q2.c6 — atmospheric ceiling vs PV ceiling: q2 still 10-100× lower than q1 (not 1,000-10,000× as I originally claimed). Atmospheric chemistry remains the binding constraint, but by a smaller margin.

What does NOT change qualitatively

• Earth chemical-rocket throughput is atmospheric-chemistry-bound below the solar-PV-area ceiling.
• Cosmic Tt/yr LEO is unreachable from Earth chemical, primarily on atmospheric grounds (with PV-area as secondary constraint).
• Methalox is dramatically lower per-launch atmospheric impact than the current mixed fuel mix (no Cl from SRMs, much less BC, no alumina from combustion).
• Reentry NOx and reentry alumina are propellant-independent — they come from atmospheric N₂ dissociation and spacecraft material ablation during reentry heating.
• The Moon's architectural necessity for cosmic-scale throughput is established by q2 (atmospheric) + q1 (solar-PV) jointly.

Next action

Patch pass-02-calc.md with the corrected reentry NOx factor + Larson 2017 anchor + revised ceiling. Then reconcile (now needs Larson 2017 fetched as a source).

leaf: q2-earth-atmospheric-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Response to Codex audit on q2 pass-02-calc

Codex caught four high-severity issues plus four medium. Most are accepted. The qualitative conclusion (atmospheric chemistry binds well before solar-PV area) holds, but the quantitative ceiling moves up by ~1-2 orders of magnitude.

Accepted — substantive corrections

• Reentry NOx 5 kg/kg-reentered (HIGH, contradicted): Codex cites Larson 2017, Vliex et al. 2024, Ryan 2022 all using ~17.5% of reentered mass as NOx (0.175 kg/kg) for reusable spacecraft, with 40% for discarded objects. My 5 kg/kg was ~30× too high. Per Block-3 Starship launch at 100 t reentered: 17.5 t NOx (not 500 t).

• Catastrophic ozone scaling (HIGH, contradicted): Larson 2017 modeled 10⁵-10⁶ launches/yr with hydrogen propellant and found ~0.5% global ozone loss at 10⁵ and ~11 DU (~3-4%) at 10⁶ launches/yr — not the "30-50% ozone loss" I extrapolated. NOx and HOx-cycle catalytic destruction saturates at high concentrations rather than scaling linearly with emissions. My linear extrapolation of Revell 2025 was wrong.

• Revell 2025 extrapolation (HIGH, unsupported): Revell's depletion is driven by Cl (from SRMs) and BC (from kerosene), NOT by NOx. Revell explicitly notes rocket-emitted NOx is insignificant in their setup AND they omit reentry NOx. Linear scaling from Revell to a methalox-only future at 10⁵-10⁶ launches/yr is unjustified.

• NOx perturbation table internal inconsistency (HIGH): my table said "5% NOx perturbation at ~90 launches/yr" but if 250 launches → 140%, then 5% binds at ~9 launches/yr. Math error in my output.

• NOx 20 g/kg propellant (MEDIUM): newer altitude-resolved inventories show NOx EI is altitude-dependent: 33 g/kg at ground falling to <1 g/kg above 14 km. The bulk-stratospheric methalox NOx contribution is much smaller than I claimed.

• NOx residence-time model (MEDIUM): NOx is short-lived in the lower stratosphere (days-weeks), not 5 years. My 5-year box-model accumulation is dimensionally wrong for NOx.

• Stratospheric injection fraction (MEDIUM): Vliex et al. 2024 splits propellant burn ~20-29% in stratosphere, ~10-19% in mesosphere, balance higher up — not the 50% I assumed.

• Combustion model (MEDIUM): fuel-rich methalox produces CO + H₂ intermediates with afterburning, not "5% CO2 deficit + 5% unburned CH4." H₂O figure stays close; CO2 figure may be lower.

Recomputed q2 ceiling

Anchoring on Larson 2017 (high-cadence H2-propellant scenarios, the closest analog to a methalox-dominant future since both lack chlorine and produce dominant H2O + NOx perturbations):

Recomputed atmospheric ceiling: ~10⁶-10⁷ launches/yr (100 Mt/yr to 1 Gt/yr LEO) — about 1-2 OOM higher than my original Mt/yr estimate, but still 10-100× lower than q1's solar-PV ceiling at ~100 Gt/yr LEO.

The qualitative conclusion holds: atmospheric chemistry is the load-bearing Earth-side constraint at any post-current cadence. Cosmic Tt/yr (10¹² t/yr = 10¹⁰ launches/yr) is still unreachable from Earth chemical — atmospheric chemistry binds by 3-4 OOM at cosmic scale.

Reentry alumina ceiling (Murphy 2023 / Maloney 2025) stays at ~10⁵-10⁶ launches/yr — but largely driven by aluminum-skinned payloads, not vehicle propellant. With non-Al satellite materials and reusable vehicles, this ceiling is partly architecture-dependent.

Disputed / clarified

• Methalox vs fossil/synthetic methane (MED, partial): my original claim was specifically about synthetic-vs-fossil methane combustion being chemically identical. Codex right that it could be misread as methalox-vs-current-fuel-mix (which is dramatically different — methalox eliminates Cl, alumina, and most BC). CLARIFY: synthetic-vs-fossil-methane is identical combustion; methalox-vs-kerosene-SRM is much cleaner. Both true; readers need both.

What changes in claims.yaml

• q2.c1 — per-launch emissions: reentry NOx corrected from 500 t → 17.5 t (factor 30 lower). H2O figure ~1,400 t stays. NOx propellant figure adjusted with altitude dependence.
• q2.c2 — atmospheric ceiling: shift "disruptive" from 10⁵ launches/yr to 10⁶, "catastrophic" from 10⁶-10⁷ to 10⁷-10⁸.
• q2.c3 — reentry NOx dominant: still load-bearing but with corrected magnitude. At Larson 17.5%, reentry NOx ≈ propellant NOx for methalox at typical reentry mass.
• q2.c5 — Tt/yr still unreachable: holds with corrected numbers. Tt/yr cadence = 10¹⁰ launches/yr, ~3-4 OOM above corrected catastrophic ceiling.
• q2.c6 — atmospheric ceiling vs PV ceiling: q2 still 10-100× lower than q1 (not 1,000-10,000× as I originally claimed). Atmospheric chemistry remains the binding constraint, but by a smaller margin.

What does NOT change qualitatively

• Earth chemical-rocket throughput is atmospheric-chemistry-bound below the solar-PV-area ceiling.
• Cosmic Tt/yr LEO is unreachable from Earth chemical, primarily on atmospheric grounds (with PV-area as secondary constraint).
• Methalox is dramatically lower per-launch atmospheric impact than the current mixed fuel mix (no Cl from SRMs, much less BC, no alumina from combustion).
• Reentry NOx and reentry alumina are propellant-independent — they come from atmospheric N₂ dissociation and spacecraft material ablation during reentry heating.
• The Moon's architectural necessity for cosmic-scale throughput is established by q2 (atmospheric) + q1 (solar-PV) jointly.

Next action

Patch pass-02-calc.md with the corrected reentry NOx factor + Larson 2017 anchor + revised ceiling. Then reconcile (now needs Larson 2017 fetched as a source).

leaf: q2-earth-atmospheric-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Response to Codex audit on q2 pass-02-calc

Codex caught four high-severity issues plus four medium. Most are accepted. The qualitative conclusion (atmospheric chemistry binds well before solar-PV area) holds, but the quantitative ceiling moves up by ~1-2 orders of magnitude.

Accepted — substantive corrections

• Reentry NOx 5 kg/kg-reentered (HIGH, contradicted): Codex cites Larson 2017, Vliex et al. 2024, Ryan 2022 all using ~17.5% of reentered mass as NOx (0.175 kg/kg) for reusable spacecraft, with 40% for discarded objects. My 5 kg/kg was ~30× too high. Per Block-3 Starship launch at 100 t reentered: 17.5 t NOx (not 500 t).

• Catastrophic ozone scaling (HIGH, contradicted): Larson 2017 modeled 10⁵-10⁶ launches/yr with hydrogen propellant and found ~0.5% global ozone loss at 10⁵ and ~11 DU (~3-4%) at 10⁶ launches/yr — not the "30-50% ozone loss" I extrapolated. NOx and HOx-cycle catalytic destruction saturates at high concentrations rather than scaling linearly with emissions. My linear extrapolation of Revell 2025 was wrong.

• Revell 2025 extrapolation (HIGH, unsupported): Revell's depletion is driven by Cl (from SRMs) and BC (from kerosene), NOT by NOx. Revell explicitly notes rocket-emitted NOx is insignificant in their setup AND they omit reentry NOx. Linear scaling from Revell to a methalox-only future at 10⁵-10⁶ launches/yr is unjustified.

• NOx perturbation table internal inconsistency (HIGH): my table said "5% NOx perturbation at ~90 launches/yr" but if 250 launches → 140%, then 5% binds at ~9 launches/yr. Math error in my output.

• NOx 20 g/kg propellant (MEDIUM): newer altitude-resolved inventories show NOx EI is altitude-dependent: 33 g/kg at ground falling to <1 g/kg above 14 km. The bulk-stratospheric methalox NOx contribution is much smaller than I claimed.

• NOx residence-time model (MEDIUM): NOx is short-lived in the lower stratosphere (days-weeks), not 5 years. My 5-year box-model accumulation is dimensionally wrong for NOx.

• Stratospheric injection fraction (MEDIUM): Vliex et al. 2024 splits propellant burn ~20-29% in stratosphere, ~10-19% in mesosphere, balance higher up — not the 50% I assumed.

• Combustion model (MEDIUM): fuel-rich methalox produces CO + H₂ intermediates with afterburning, not "5% CO2 deficit + 5% unburned CH4." H₂O figure stays close; CO2 figure may be lower.

Recomputed q2 ceiling

Anchoring on Larson 2017 (high-cadence H2-propellant scenarios, the closest analog to a methalox-dominant future since both lack chlorine and produce dominant H2O + NOx perturbations):

Recomputed atmospheric ceiling: ~10⁶-10⁷ launches/yr (100 Mt/yr to 1 Gt/yr LEO) — about 1-2 OOM higher than my original Mt/yr estimate, but still 10-100× lower than q1's solar-PV ceiling at ~100 Gt/yr LEO.

The qualitative conclusion holds: atmospheric chemistry is the load-bearing Earth-side constraint at any post-current cadence. Cosmic Tt/yr (10¹² t/yr = 10¹⁰ launches/yr) is still unreachable from Earth chemical — atmospheric chemistry binds by 3-4 OOM at cosmic scale.

Reentry alumina ceiling (Murphy 2023 / Maloney 2025) stays at ~10⁵-10⁶ launches/yr — but largely driven by aluminum-skinned payloads, not vehicle propellant. With non-Al satellite materials and reusable vehicles, this ceiling is partly architecture-dependent.

Disputed / clarified

• Methalox vs fossil/synthetic methane (MED, partial): my original claim was specifically about synthetic-vs-fossil methane combustion being chemically identical. Codex right that it could be misread as methalox-vs-current-fuel-mix (which is dramatically different — methalox eliminates Cl, alumina, and most BC). CLARIFY: synthetic-vs-fossil-methane is identical combustion; methalox-vs-kerosene-SRM is much cleaner. Both true; readers need both.

What changes in claims.yaml

• q2.c1 — per-launch emissions: reentry NOx corrected from 500 t → 17.5 t (factor 30 lower). H2O figure ~1,400 t stays. NOx propellant figure adjusted with altitude dependence.
• q2.c2 — atmospheric ceiling: shift "disruptive" from 10⁵ launches/yr to 10⁶, "catastrophic" from 10⁶-10⁷ to 10⁷-10⁸.
• q2.c3 — reentry NOx dominant: still load-bearing but with corrected magnitude. At Larson 17.5%, reentry NOx ≈ propellant NOx for methalox at typical reentry mass.
• q2.c5 — Tt/yr still unreachable: holds with corrected numbers. Tt/yr cadence = 10¹⁰ launches/yr, ~3-4 OOM above corrected catastrophic ceiling.
• q2.c6 — atmospheric ceiling vs PV ceiling: q2 still 10-100× lower than q1 (not 1,000-10,000× as I originally claimed). Atmospheric chemistry remains the binding constraint, but by a smaller margin.

What does NOT change qualitatively

• Earth chemical-rocket throughput is atmospheric-chemistry-bound below the solar-PV-area ceiling.
• Cosmic Tt/yr LEO is unreachable from Earth chemical, primarily on atmospheric grounds (with PV-area as secondary constraint).
• Methalox is dramatically lower per-launch atmospheric impact than the current mixed fuel mix (no Cl from SRMs, much less BC, no alumina from combustion).
• Reentry NOx and reentry alumina are propellant-independent — they come from atmospheric N₂ dissociation and spacecraft material ablation during reentry heating.
• The Moon's architectural necessity for cosmic-scale throughput is established by q2 (atmospheric) + q1 (solar-PV) jointly.

Next action

Patch pass-02-calc.md with the corrected reentry NOx factor + Larson 2017 anchor + revised ceiling. Then reconcile (now needs Larson 2017 fetched as a source).

leaf: q2-earth-atmospheric-ceiling pass: write started: 2026-06-09T03:50:00Z ended: 2026-06-09T04:15:00Z researcher: claude-opus-4-7

Earth's atmospheric ceiling on chemical-rocket throughput

Why this question matters

q1 established that industrial inputs and solar PV area don't bind Earth chemical-rocket throughput until ~100 Gt/yr LEO — comfortably above current launch rates by many orders of magnitude. But the atmosphere is a finite chemical reservoir, and rocket exhaust + reentry products accumulate in the stratosphere with multi-year residence times. The question is: at what launch cadence does Earth's atmospheric chemistry break? Is that ceiling above or below q1's industrial-input ceiling?

Where this fits

q2 is the Earth-side companion to q1. Together they bound the upper limit of Earth chemical-rocket throughput. The lower of the two is the actually-binding constraint. q9 synthesis combines both with q3 (propulsion-class destination reachability) and q5/q6/q8a/q8b (lunar capability) to answer the report's root question.

Headline answer

Atmospheric chemistry binds Earth chemical-rocket throughput at approximately 100 Mt/yr to 1 Gt/yr LEO mass-flux — 10-100× below the q1 solar-PV ceiling. [q2.c2, q2.c6] Anchored on Larson 2017's chemistry-climate modeling of high-cadence hydrogen-propellant launches (the closest peer-reviewed analog to a future methalox-dominant regime), the tiers are:

At cosmic Tt/yr LEO (10¹² t/yr = 10¹⁰ launches/yr Block-3 reusable), atmospheric chemistry binds by 3-4 orders of magnitude beyond what any chemistry-climate model has explored. [q2.c5] The Moon's architectural necessity at cosmic Tt/yr scale is jointly established by q1 (solar PV area saturates at 100 Gt/yr) and q2 (atmospheric chemistry saturates 10-100× lower) — q2 binds first.

Confidence: high on the qualitative ordering (atmospheric binds before industrial); medium-high on the specific Mt-Gt/yr ceiling numbers (sensitive to the linear vs saturated extrapolation from Larson's modeled cadences).

Three pathways through the atmosphere

Per Block 3 Starship launch (methalox), combustion + reentry deposit into the stratosphere approximately:

• ~700 t H₂O stratospheric injection per launch (from CH₄ + 2 O₂ → CO₂ + 2 H₂O combustion; ~20-29% of products reach above tropopause) [q2.c1]
• ~600 t CO₂ stratospheric injection — small compared to atmospheric reservoir but accumulates
• ~18 t NOx from reentry heat (17.5% of reentered mass per Larson 2017 / Vliex 2024) [q2.c3]
• ~10-50 t NOx from propellant combustion (altitude-dependent emission index; lower in upper stratosphere)
• ~3 t black carbon — much less than the 35 g/kg of kerosene-fuelled rockets; methalox is much cleaner
• Reentry alumina from payloads — 10 kt/yr at the 2040 60k-LEO-satellite projection (Maloney 2025) [q2.c7]

The three independent atmospheric impact pathways:

Direct ozone destruction via NOx + HOx catalytic cycles. Reentry NOx and stratospheric H₂O drive ozone loss via Cl-free chemistry (since methalox has no chlorine). The catalytic cycles saturate at high concentrations — Larson 2017's direct modeling shows ~0.5% global ozone loss at 10⁵ launches/yr and ~11 DU (~3-4%) at 10⁶ launches/yr, far less than naive linear extrapolation from current low-cadence data would predict.

Stratospheric water-vapor perturbation. Methalox combustion deposits H₂O into the stratosphere where it has multi-year residence and drives HOx-catalyzed ozone loss + warming via greenhouse forcing. At 10⁶ launches/yr (100 Mt/yr LEO), steady-state stratospheric H₂O is ~25% above natural; at 10⁷ launches/yr the stratosphere has substantially anthropogenic water vapor.

Reentry metal aerosol. Murphy 2023 found ~10% of stratospheric sulfuric acid particles already contain spacecraft-origin metals. Maloney 2025 projects 10 kt/yr alumina at 2040 megaconstellation cadence with +1.5°C polar mesospheric warming and weakened polar vortex. This is largely independent of propellant chemistry and partly architecture-dependent (aluminum-skinned satellites are the main contributor).

Why the Handmer solar-abundance regime does not help

The Handmer/Terraform synthetic-methane thesis (from q1) makes propellant supply effectively unlimited from solar PV + electrolysis + Sabatier. But synthetic methane is chemically identical to fossil methane. Combustion produces the same CO₂ + H₂O + NOx + BC regardless of whether the methane came from a gas well or from the sky. [q2.c4]

The solar-abundance regime helps Earth's carbon cycle (no net new CO₂ to the atmosphere over the synthesis-combustion cycle) but does NOT relax q2's stratospheric chemistry ceiling. Per-launch atmospheric perturbation is the same.

This is the critical distinction between q1 (where solar abundance softens the ceiling) and q2 (where it doesn't): industrial inputs are willingness-to-scale, but atmospheric chemistry is fundamentally physical. Pollutants accumulate regardless of where the rocket fuel came from. [q2.c8]

What this means for the Moon

The Moon's architectural necessity for cosmic Tt/yr LEO mass throughput is established jointly by q1 + q2, with q2 binding first. At the more modest 100 Mt/yr - 1 Gt/yr LEO scale (where serious solar-system industrialisation actually wants to operate — building space-based solar, large orbital infrastructure, deep-space mission staging), q2 atmospheric chemistry is the dominant Earth-side constraint, binding 10-100× below q1's solar-PV ceiling.

Mass driver launch from the lunar surface bypasses Earth's atmosphere entirely. There is no q2-equivalent ceiling on the Moon side because the lunar exosphere is ~10⁻¹² of Earth's atmospheric mass and cannot meaningfully accumulate pollutants in the same way. This is the architectural argument that runs through the report.

Confidence per claim

Limitations and known omissions

• Larson 2017 was retrieved via Codex audit citation, not independently inspected; future audit re-pass should fetch the full PDF directly.
• NASA TM-20240013276 Impact of Spaceflight on Earth's Atmosphere failed to fetch cleanly (PDF binary issue); should be added via alternative retrieval.
• Vliex et al. 2024 (gridded inventory in Scientific Data) cited via Codex but not separately extracted; useful for altitude-resolved NOx profile in re-pass.
• NOx catalytic-cycle saturation is asserted from Larson 2017 but not derived from first principles here; q2 should be re-modeled with explicit catalytic-cycle chemistry to give precise ceilings.
• Polar vs equatorial geographic concentration of ozone effects is mentioned but not separately bounded.
• Long-term stratospheric H₂O climate forcing at multi-decade timescale beyond steady-state residence is not modeled.
• Plume-scale chemistry (very near rocket nozzle, on minute timescales) interacts with bulk stratosphere chemistry in ways our box-model doesn't capture.

What changes if the answer flips

The q2 ceiling could move higher under three scenarios:

1. Below-stratosphere launches — if launch architecture were redesigned so combustion + reentry occurred only below the tropopause (~12 km), atmospheric chemistry would not bind. This requires fundamentally different launch vehicles (no orbital insertion above the troposphere) — physically impossible for chemical rockets reaching LEO at 200+ km.

2. Active stratospheric remediation — engineered removal of stratospheric NOx, H₂O, alumina at the same rate they're injected. Theoretically possible at vast cost; would essentially require a global stratospheric scrubbing system. Not a current technology and probably never economic.

3. Atmospheric monitoring catches saturation effects much earlier than Larson 2017's model predicts, pushing the ceiling down to 10⁵-10⁶ launches/yr. This would mean q2 binds harder, not softer — strengthening the Moon-necessity argument.

The ceiling could move lower if reentry NOx is higher than Larson's 17.5%, or if polar vortex effects from alumina destabilize ozone chemistry more than current models suggest.

The atmospheric ceiling is genuinely fundamental

This is the load-bearing finding of q2 and the most important difference from q1. Atmospheric chemistry is not willingness-to-scale; it is physically constrained by:

• Atmospheric mass (~5,150 Tt, finite)
• Stratospheric residence times (multi-year for H₂O and aerosols, days-weeks for NOx)
• Catalytic-cycle chemistry (NOx, HOx, ClOx coupling)
• Reentry energy deposition (kinetic energy → NO formation independent of mass)

There is no architecture redesign within the chemical-rocket regime that relaxes these constraints. The Moon's value at cosmic-scale throughput rests primarily on q2.

overall: verdict: weak summary: 'Near-term arithmetic is mostly internally consistent, but the pass is not genuinely first-principles and the high-scale ceiling claims contain major inconsistencies. The LOX-first ordering is plausible under the stated assumptions, while the 10^6-10^8 t/yr and Tt/yr conclusions are under-derived.'

findings:

• target: 'method.first_principles' target_kind: method verdict: fail reason: 'Many key denominators are borrowed, not derived: US/global oxygen production, US gas consumption, stainless production, electricity generation, Raptor production target, pad count, and Baytown-class ASU capacity. That is acceptable as assumptions, but contradicts the pass framing.' severity: high

• target: 'goal.5_50_100_ceiling_coverage' target_kind: completeness verdict: fail reason: 'The stated goal requires 5%, 50%, and 100% ceilings for each input. Several inputs omit 50% cases, and engines/pads are not expressed as fractions of a defined supply denominator.' severity: high

• target: 'q1.c1' target_kind: claim verdict: partial reason: 'The LOX arithmetic is correct given 5,650 t propellant, O/F 3.6, and 10.3 Mt/yr oxygen supply. But the denominator is industrial oxygen, not necessarily merchant liquid oxygen capacity, storage, transport, or load-rate capacity, so the claimed LOX ceiling may be overstated.' severity: medium

• target: 'assumption.3' target_kind: assumption verdict: partial reason: 'The 3.6:1 methalox mixture ratio is plausible for Raptor-like engines, but it is not stoichiometric; stoichiometric CH4/O2 is 4.0:1 by mass. The numerical impact is small, but the stated physical basis is wrong.' severity: low

• target: 'step2.ng_conversion_comment' target_kind: calculation verdict: partial reason: 'The computed 50,000 scf/t methane is approximately right, but the comment gives garbled units: it should be about 1,400 m3 gas per tonne methane divided by 0.0283 m3/scf, not 1.4 m3/t times 1.42 m3/scf.' severity: low

• target: 'q1.c2' target_kind: claim verdict: weak reason: 'The 2,564 launches/yr calculation follows from 1,000 engines/yr and 39 engines amortized over 100 flights. But both the 1,000-engine target and 100-flight life are borrowed/aspirational, and engine bill-of-material constraints are not analyzed.' severity: medium

• target: 'q1.c3' target_kind: claim verdict: partial reason: 'The pad arithmetic is correct for 50-100 pads at 1-3 launches/day. The claim is still assumption-driven and omits pad refurbishment, deluge water, GSE, local range constraints, methane/LOX loading infrastructure, and site geography.' severity: medium

• target: 'q1.c4' target_kind: claim verdict: partial reason: 'Natural-gas feedstock is correctly shown as non-binding relative to LOX under the stated assumptions. However, rocket-grade LCH4 purification, liquefaction capacity, storage, boiloff, and liquefaction electricity are not included.' severity: medium

• target: 'q1.c5' target_kind: claim verdict: partial reason: 'The reusable steel arithmetic is internally correct if a 275 t stack lasts 100 flights. It omits fleet buildout, attrition, scrap/yield, specific stainless alloy availability, and steel in pads, tanks, ASUs, and GSE.' severity: low

• target: 'q1.c6' target_kind: claim verdict: supports reason: 'The ASU electricity calculation is dimensionally correct: 300 kWh/t times 4,422 t LOX gives about 1.33 GWh/launch, making grid electricity non-binding before LOX capacity. Caveat: this excludes methane liquefaction and some liquid-product/storage losses.' severity: low

• target: 'q1.c7' target_kind: claim verdict: contradicted reason: 'The high-end 100 Mt/yr case equals about 1,000,000 launches/yr. At 100-flight reuse that needs about 390,000 engines/yr, not 10^7, and at 1,000 launches/pad-yr it needs about 1,000 pads, not 10^5.' severity: high

• target: 'q1.c8' target_kind: claim verdict: weak reason: 'The Tt/yr arithmetic is mostly correct, but the "4-5 OOM below" statement contradicts the stated 10^6-10^8 t/yr ceiling, which is only 1-3 OOM below 10^9 t/yr. The impossibility conclusion is plausible but not established by the presented derivation.' severity: high

notes:

• issue: 'Helium is omitted. Starship may use autogenous pressurization, but purge, testing, leak-checking, pneumatic, and ground-system helium demand should be explicitly bounded because helium supply is small and strategically constrained.' severity: medium
• issue: 'Hydrogen is out of scope for a pure methalox Starship analysis, but not for an "Earth chemical rockets" ceiling. Hydrolox alternatives require LH2 production, liquefaction electricity, insulation, boiloff, and large-volume tankage.' severity: medium
• issue: 'Cryogenic insulation, vacuum-jacketed piping, perlite/MLI, tank farms, pumps, valves, and high-flow loading hardware are missing from the industrial-input list.' severity: medium
• issue: 'Avionics, IMUs, flight computers, sensors, batteries, harnessing, and space-rated electronics are omitted. Probably not binding near term, but relevant at 10^5-10^7 launches/yr.' severity: low
• issue: 'Copper and high-temperature alloys are omitted from the Raptor constraint. Combustion chambers, turbomachinery, wiring, power systems, and pad electrical infrastructure could become material at aircraft- or auto-scale engine production.' severity: medium
• issue: 'The analysis mixes US and global denominators: US LOX/grid/gas are used for some ceilings, global oxygen for others, and cosmic-scale ASU power is compared to US electricity despite being framed as an Earth-scale limit.' severity: medium
• issue: 'Ground tests, static fires, chilldown losses, boiloff, scrubbed launches, reserve propellant, and production rejects are excluded, so per-launch propellant and engine demand are optimistic.' severity: medium

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

overall: verdict: weak summary: 'Near-term arithmetic is mostly internally consistent, but the pass is not genuinely first-principles and the high-scale ceiling claims contain major inconsistencies. The LOX-first ordering is plausible under the stated assumptions, while the 10^6-10^8 t/yr and Tt/yr conclusions are under-derived.'

findings:

• target: 'method.first_principles' target_kind: method verdict: fail reason: 'Many key denominators are borrowed, not derived: US/global oxygen production, US gas consumption, stainless production, electricity generation, Raptor production target, pad count, and Baytown-class ASU capacity. That is acceptable as assumptions, but contradicts the pass framing.' severity: high

• target: 'goal.5_50_100_ceiling_coverage' target_kind: completeness verdict: fail reason: 'The stated goal requires 5%, 50%, and 100% ceilings for each input. Several inputs omit 50% cases, and engines/pads are not expressed as fractions of a defined supply denominator.' severity: high

• target: 'q1.c1' target_kind: claim verdict: partial reason: 'The LOX arithmetic is correct given 5,650 t propellant, O/F 3.6, and 10.3 Mt/yr oxygen supply. But the denominator is industrial oxygen, not necessarily merchant liquid oxygen capacity, storage, transport, or load-rate capacity, so the claimed LOX ceiling may be overstated.' severity: medium

• target: 'assumption.3' target_kind: assumption verdict: partial reason: 'The 3.6:1 methalox mixture ratio is plausible for Raptor-like engines, but it is not stoichiometric; stoichiometric CH4/O2 is 4.0:1 by mass. The numerical impact is small, but the stated physical basis is wrong.' severity: low

• target: 'step2.ng_conversion_comment' target_kind: calculation verdict: partial reason: 'The computed 50,000 scf/t methane is approximately right, but the comment gives garbled units: it should be about 1,400 m3 gas per tonne methane divided by 0.0283 m3/scf, not 1.4 m3/t times 1.42 m3/scf.' severity: low

• target: 'q1.c2' target_kind: claim verdict: weak reason: 'The 2,564 launches/yr calculation follows from 1,000 engines/yr and 39 engines amortized over 100 flights. But both the 1,000-engine target and 100-flight life are borrowed/aspirational, and engine bill-of-material constraints are not analyzed.' severity: medium

• target: 'q1.c3' target_kind: claim verdict: partial reason: 'The pad arithmetic is correct for 50-100 pads at 1-3 launches/day. The claim is still assumption-driven and omits pad refurbishment, deluge water, GSE, local range constraints, methane/LOX loading infrastructure, and site geography.' severity: medium

• target: 'q1.c4' target_kind: claim verdict: partial reason: 'Natural-gas feedstock is correctly shown as non-binding relative to LOX under the stated assumptions. However, rocket-grade LCH4 purification, liquefaction capacity, storage, boiloff, and liquefaction electricity are not included.' severity: medium

• target: 'q1.c5' target_kind: claim verdict: partial reason: 'The reusable steel arithmetic is internally correct if a 275 t stack lasts 100 flights. It omits fleet buildout, attrition, scrap/yield, specific stainless alloy availability, and steel in pads, tanks, ASUs, and GSE.' severity: low

• target: 'q1.c6' target_kind: claim verdict: supports reason: 'The ASU electricity calculation is dimensionally correct: 300 kWh/t times 4,422 t LOX gives about 1.33 GWh/launch, making grid electricity non-binding before LOX capacity. Caveat: this excludes methane liquefaction and some liquid-product/storage losses.' severity: low

• target: 'q1.c7' target_kind: claim verdict: contradicted reason: 'The high-end 100 Mt/yr case equals about 1,000,000 launches/yr. At 100-flight reuse that needs about 390,000 engines/yr, not 10^7, and at 1,000 launches/pad-yr it needs about 1,000 pads, not 10^5.' severity: high

• target: 'q1.c8' target_kind: claim verdict: weak reason: 'The Tt/yr arithmetic is mostly correct, but the "4-5 OOM below" statement contradicts the stated 10^6-10^8 t/yr ceiling, which is only 1-3 OOM below 10^9 t/yr. The impossibility conclusion is plausible but not established by the presented derivation.' severity: high

notes:

• issue: 'Helium is omitted. Starship may use autogenous pressurization, but purge, testing, leak-checking, pneumatic, and ground-system helium demand should be explicitly bounded because helium supply is small and strategically constrained.' severity: medium
• issue: 'Hydrogen is out of scope for a pure methalox Starship analysis, but not for an "Earth chemical rockets" ceiling. Hydrolox alternatives require LH2 production, liquefaction electricity, insulation, boiloff, and large-volume tankage.' severity: medium
• issue: 'Cryogenic insulation, vacuum-jacketed piping, perlite/MLI, tank farms, pumps, valves, and high-flow loading hardware are missing from the industrial-input list.' severity: medium
• issue: 'Avionics, IMUs, flight computers, sensors, batteries, harnessing, and space-rated electronics are omitted. Probably not binding near term, but relevant at 10^5-10^7 launches/yr.' severity: low
• issue: 'Copper and high-temperature alloys are omitted from the Raptor constraint. Combustion chambers, turbomachinery, wiring, power systems, and pad electrical infrastructure could become material at aircraft- or auto-scale engine production.' severity: medium
• issue: 'The analysis mixes US and global denominators: US LOX/grid/gas are used for some ceilings, global oxygen for others, and cosmic-scale ASU power is compared to US electricity despite being framed as an Earth-scale limit.' severity: medium
• issue: 'Ground tests, static fires, chilldown losses, boiloff, scrubbed launches, reserve propellant, and production rejects are excluded, so per-launch propellant and engine demand are optimistic.' severity: medium

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

leaf: q1-earth-industrial-ceiling pass: write started: 2026-06-08T23:15:00Z ended: 2026-06-08T23:45:00Z researcher: claude-opus-4-7

Earth Industrial Inputs — Where the Ceiling Sits

Why this question matters

The cosmic-scale-throughput question — can humanity move very large quantities of mass into orbit and onward across the solar system — only matters if we know where Earth-launched chemical rockets cap out. The cap is set somewhere: LOX production, methane supply, engine cycle-life, pad turnaround, atmosphere. This leaf identifies which of those binds first, and at what flux.

Where this fits

q1 is the industrial-supply-side ceiling on Earth-launched chemical-rocket throughput. Sibling q2 covers the atmospheric-chemistry ceiling. q3 maps LEO mass-flux to interplanetary destinations by propulsion class. Synthesis q9 integrates the Earth-side and Moon-side ceilings.

Headline answer

LOX binds first. [q1.c1] Soft ceiling at ~120 Block-3 Starship launches per year (5% of current US oxygen production); hard ceiling at ~2,330 launches per year (100% of US O₂), equivalent to roughly 12 to 230 kt to LEO per year without building dedicated rocket-grade LOX capacity. With aggressive industrial scaling (dedicated LOX industry ~15× current global O₂ supply, ~2,700 launch pads, ~390,000 engines/year, ~130 GW dedicated ASU electricity), the Earth chemical-rocket throughput ceiling sits at approximately 1-100 Mt/year to LEO [q1.c8] — one to three orders of magnitude below the Tt/year cosmic scale the parent report is calibrated to [q1.c9]. Earth chemical rockets cannot deliver Tt/yr to LEO under any plausible industrial scaling.

Confidence: high on the ordering and the Tt/yr-unreachable conclusion; medium-high on the specific kt/yr ceiling numbers.

Binding-input ordering

The ordering — which input binds first as Starship cadence climbs — is the load-bearing structural finding. Methane gets the cinematic attention because Starship is visibly burning LNG, but supply-chain arithmetic puts it nowhere near the binding constraint.

LOX binds first because methalox burns at 3.6:1 LOX-to-methane by mass (Raptor's operating ratio, slightly fuel-rich relative to the 4.0:1 stoichiometric ratio). Most of the propellant is oxygen, and the supply of cryogenic-grade oxygen is pegged to bulk air-separation-unit capacity. Methane is abundant — pipeline-grade US natural gas at 33.5 Tcf/yr would support over half a million Block-3 launches per year before saturating. The asymmetry is structural: oxygen is the supply chain that has to scale.

Engine production is the second-binding constraint under realistic reuse assumptions. SpaceX's stated Raptor production target is ~1,000 engines/year; at 39 engines per stack and 100-flight aspirational reuse, that supports ~2,600 launches/year. If demonstrated reuse turns out to be 10 flights rather than 100, the ceiling drops to ~260 launches/year, putting it inside the LOX-binding band. Musk has stated a 10,000-ships-per-year production aspiration; at any reuse cadence above ~3 flights/ship the binding constraint shifts back to LOX. [q1.c10]

Pads, methane, steel for reusable operations, and ASU electricity all sit above the LOX-and-engine ceilings, in the 10⁴-10⁶ launches/yr regime. They become binding only after several rounds of dedicated industrial scale-up.

Where the ceiling sits at maturity

With civilizational-scale industrial mobilisation — a dedicated rocket-grade LOX industry, ASU build-outs measured in thousands of Baytown-class facilities, a global pad network at order-of-magnitude expansion from current sites, engine production at aircraft-industry scale — the ceiling lands somewhere in the 1-100 Mt/yr LEO range. [q1.c8]

The 100 Mt/yr line sits at the upper bound of plausible Earth chemical-rocket infrastructure. 1,350 ASUs at ~$850M each represents ~$1.1 trillion of new capital; 2,740 pads is roughly 100× current active launch sites globally; 390,000 engines per year is ~400× current Raptor production target. Each axis represents an order-of-magnitude scale-up from any current industrial system in that category. Reaching this ceiling is mechanically possible — none of the steps violates physics or known engineering — but each requires sustained industrial mobilisation on a scale not recently undertaken.

What this means for the cosmic-scale question

The throughput target the parent report is calibrated to (Tt/yr to LEO, Dyson-relevant) is 10⁹ t/yr. The chemical ceiling derived here is 10⁶-10⁸ t/yr. The gap is one to three orders of magnitude, with the lower end of the gap reachable only after civilizational-scale mobilisation. [q1.c9]

Tt/yr from Earth chemical alone would require:

• LOX industry ~50× to ~500× current global O₂ supply
• ~10⁴ launch pads (against the historical maximum of ~30-50 simultaneously-active sites)
• ~10⁷ engines per year — roughly 100× the global engine production across all commercial aviation
• ~1 TW dedicated air-separation electricity, exceeding current US generation capacity

None of these is a physics problem; each is a scale problem. Humanity's largest current industries — global steel ~2 Gt/yr, global oil ~5 Gt/yr, global electricity ~30,000 TWh/yr — were built over a century with multiple complementary supply chains. Reaching Tt/yr LEO through Earth chemical rockets requires growing a rocket-oxygen industry to roughly the scale of several of those existing industries combined, in a coordinated reorganisation around a single product. The Tt/yr cosmic-scale demand cannot be served by Earth chemical-rocket launch under any plausible scaling. [q1.c9]

This sets the upper bound on Earth alone. Whether the Moon channel can reach above this bound is the question for the rest of the report; q9 makes the integration.

Confidence per finding

Limitations

A handful of industrial inputs are out of scope here and worth surfacing for the audit pass:

• Helium supply is a real but unbounded constraint. Starship is autogenous-pressurized at flight, but ground-system helium (purge, leak testing, GSE pneumatics) is non-trivial, and global helium production is small (~30 kt/yr) and strategically concentrated.
• Cryogenic insulation, vacuum-jacketed plumbing, multi-layer insulation, pad propellant farms, transfer infrastructure are not bounded here.
• Avionics, IMUs, flight computers, sensors, batteries are non-binding at the cadences analysed but worth flagging at 10⁵+ launches/yr.
• Copper and high-temperature alloys become material at aircraft-scale engine production for the Raptor bill-of-materials.
• Merchant LOX vs bulk industrial O₂: the 10.3 Mt/yr figure is total US oxygen production. Merchant liquid oxygen capacity (storage, transport, rail/road loading) is a subset; the LOX ceiling quoted is an upper bound, and the operational rocket-LOX bottleneck is somewhat tighter.
• Ground-system overhead: per-launch propellant demand here is the vehicle load. Real ground demand is ~1.2-1.5× higher because of static fires, chilldown, boiloff, scrubs, reserves, and rejects. The ceiling numbers therefore overstate the true rocket-launch ceiling by ~20-50%.
• Hydrolox alternative: out of scope (the report is calibrated to methalox baseline). LOX-binding result is broadly invariant since LH₂ also burns with LOX at ~6-8:1 mass ratio.
• EPA tier-S primary retrieval failure: the US LOX supply figure rests on a secondary citation rather than fresh primary PDF retrieval. A future re-pass should fetch the EPA HTML or use USGS / Industrial Gas Association alternative.

All limitations are documented; none change the qualitative ordering or the categorical-infeasibility conclusion at cosmic scale.

What changes if the answer flips

The ceiling could move up substantially under three scenarios:

1. Non-chemical Earth launch matures. Launch loops (Lofstrom; 4 GW supports 750 kt/yr at $3/kg), skyhooks, laser propulsion, or nuclear thermal rockets — all out of scope for this leaf — could each lift the ceiling by orders of magnitude. None is currently TRL 5+.
2. Hydrolox + advanced ASU electrolysis chains could shift the per-launch oxygen demand profile. Marginal, not order-of-magnitude.
3. TAI-grade automation compresses the build-out timeline so that 10⁸ t/yr LEO is reached in a decade rather than a century. This does not change the ceiling itself, only the time to approach it.

None of these moves the ceiling from 10⁸ to 10⁹ t/yr without invoking either (a) a new launch technology or (b) the Moon. The Moon channel is the rest of the report.

leaf: q1-earth-industrial-ceiling pass: write-response responding_to: pass-write-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Response to Codex write-pass audit

Accepted (will patch the write)

• Headline OOM contradiction (high): I wrote "three to four orders of magnitude below" Tt/yr in the headline, then "one to three OOM" later in the same paragraph and in q1.c9. Internal contradiction. Fix: change headline to "one to three orders of magnitude" — single, consistent OOM count throughout.

• Bury-the-lead at ~256 words (medium): the motivation + where-it-fits sections push the headline past 200 words. Fix: compress motivation to ~80 words and where-it-fits to ~60 words, putting headline at ~150 words.

• "Manhattan Project / Interstate Highway / aircraft manufacturing" comparison (high unsupported): these are rhetorical flourishes not encoded in any claim. Fix: remove the sentence; keep only the quantitative substance (ASU count, pad count, engine count, ASU power).

• "Architecturally necessary" / "load-bearing handoff" (high unsupported synthesis): these are synthesis moves belonging to q9, not q1. Fix: soften to "this is the upper bound on Earth alone; q9 will determine whether the Moon channel reaches above it."

• Editorialising phrases (medium): Codex flagged "by a wide margin," "civilizational-scale," "categorical-scale infeasibility," "architecturally necessary," "enormous," "cinematic-imagination," "not a luxury," "load-bearing piece of architecture." Per anti-pattern #13, replace with quantitative substance. Fix: remove most; the few that remain ("civilizational-scale") will be re-cast as quantitative ("requires ~$1.1T new capex").

Disputed / clarified

• "Why this matters / Where it fits" not in claims (medium): the leaf-write schema (anti-pattern #12) requires a motivation + where-it-fits section before the headline. Codex flagged this as unsupported, but the spec explicitly allows context-framing here. Clarification: these sections are structural framing, not load-bearing claims; the per-claim audit applies to the body, not the framing.

• Manhattan/Interstate as rhetorical flourishes (high): ACCEPT. Removing.

• Helium / cryogenic / avionics / hydrolox / EPA-retrieval in Limitations (medium): Codex flagged these as not-in-claims. But the leaf-write schema (anti-pattern #12 step 6) requires a Limitations section. Per the source-tier schema, Limitations are known-omissions, not load-bearing claims that need to appear in claims.yaml. Clarification: these are surfaced from pass-02-calc.md "Caveats and known omissions" footer (Codex audit informed) — they belong in Limitations by design.

• Launch loop / TRL / hydrolox / TAI in "What changes if the answer flips" (medium): these are sensitivity-scenario sketches, not claims. ACCEPT in principle: should mark as type: estimate claims if cited as facts. Fix: soften to "out-of-scope for this leaf; flagged for future cross-leaf work" without specific numbers; or move the specific numbers to claims with confidence: speculative.

What changes substantively

• Headline OOM number (fix from 3-4 to 1-3).
• Compress motivation + where-it-fits to ≤200 combined words.
• Remove Manhattan Project / Interstate Highway / aircraft-manufacturing rhetorical comparisons.
• Soften "architecturally necessary" / synthesis-move framing.
• Trim editorialising verbs (anti-pattern #13).
• Add a Note: "What changes if the answer flips" section's specific numbers (Lofstrom 4 GW / 750 kt/yr / $3/kg) cite the workspace anchor extract; mark as speculative/sensitivity-only.

What does not change

• The binding-input ordering (LOX → engines → pads → ASU power → methane → steel).
• The 1-100 Mt/yr ceiling band.
• The Tt/yr-impossibility conclusion.
• The confidence labels per claim.
• The structure of the leaf-write (motivation → where-it-fits → headline → body → confidence → limitations).

overall: verdict: weak summary: 'Near-term arithmetic is mostly internally consistent, but the pass is not genuinely first-principles and the high-scale ceiling claims contain major inconsistencies. The LOX-first ordering is plausible under the stated assumptions, while the 10^6-10^8 t/yr and Tt/yr conclusions are under-derived.'

findings:

• target: 'method.first_principles' target_kind: method verdict: fail reason: 'Many key denominators are borrowed, not derived: US/global oxygen production, US gas consumption, stainless production, electricity generation, Raptor production target, pad count, and Baytown-class ASU capacity. That is acceptable as assumptions, but contradicts the pass framing.' severity: high

• target: 'goal.5_50_100_ceiling_coverage' target_kind: completeness verdict: fail reason: 'The stated goal requires 5%, 50%, and 100% ceilings for each input. Several inputs omit 50% cases, and engines/pads are not expressed as fractions of a defined supply denominator.' severity: high

• target: 'q1.c1' target_kind: claim verdict: partial reason: 'The LOX arithmetic is correct given 5,650 t propellant, O/F 3.6, and 10.3 Mt/yr oxygen supply. But the denominator is industrial oxygen, not necessarily merchant liquid oxygen capacity, storage, transport, or load-rate capacity, so the claimed LOX ceiling may be overstated.' severity: medium

• target: 'assumption.3' target_kind: assumption verdict: partial reason: 'The 3.6:1 methalox mixture ratio is plausible for Raptor-like engines, but it is not stoichiometric; stoichiometric CH4/O2 is 4.0:1 by mass. The numerical impact is small, but the stated physical basis is wrong.' severity: low

• target: 'step2.ng_conversion_comment' target_kind: calculation verdict: partial reason: 'The computed 50,000 scf/t methane is approximately right, but the comment gives garbled units: it should be about 1,400 m3 gas per tonne methane divided by 0.0283 m3/scf, not 1.4 m3/t times 1.42 m3/scf.' severity: low

• target: 'q1.c2' target_kind: claim verdict: weak reason: 'The 2,564 launches/yr calculation follows from 1,000 engines/yr and 39 engines amortized over 100 flights. But both the 1,000-engine target and 100-flight life are borrowed/aspirational, and engine bill-of-material constraints are not analyzed.' severity: medium

• target: 'q1.c3' target_kind: claim verdict: partial reason: 'The pad arithmetic is correct for 50-100 pads at 1-3 launches/day. The claim is still assumption-driven and omits pad refurbishment, deluge water, GSE, local range constraints, methane/LOX loading infrastructure, and site geography.' severity: medium

• target: 'q1.c4' target_kind: claim verdict: partial reason: 'Natural-gas feedstock is correctly shown as non-binding relative to LOX under the stated assumptions. However, rocket-grade LCH4 purification, liquefaction capacity, storage, boiloff, and liquefaction electricity are not included.' severity: medium

• target: 'q1.c5' target_kind: claim verdict: partial reason: 'The reusable steel arithmetic is internally correct if a 275 t stack lasts 100 flights. It omits fleet buildout, attrition, scrap/yield, specific stainless alloy availability, and steel in pads, tanks, ASUs, and GSE.' severity: low

• target: 'q1.c6' target_kind: claim verdict: supports reason: 'The ASU electricity calculation is dimensionally correct: 300 kWh/t times 4,422 t LOX gives about 1.33 GWh/launch, making grid electricity non-binding before LOX capacity. Caveat: this excludes methane liquefaction and some liquid-product/storage losses.' severity: low

• target: 'q1.c7' target_kind: claim verdict: contradicted reason: 'The high-end 100 Mt/yr case equals about 1,000,000 launches/yr. At 100-flight reuse that needs about 390,000 engines/yr, not 10^7, and at 1,000 launches/pad-yr it needs about 1,000 pads, not 10^5.' severity: high

• target: 'q1.c8' target_kind: claim verdict: weak reason: 'The Tt/yr arithmetic is mostly correct, but the "4-5 OOM below" statement contradicts the stated 10^6-10^8 t/yr ceiling, which is only 1-3 OOM below 10^9 t/yr. The impossibility conclusion is plausible but not established by the presented derivation.' severity: high

notes:

• issue: 'Helium is omitted. Starship may use autogenous pressurization, but purge, testing, leak-checking, pneumatic, and ground-system helium demand should be explicitly bounded because helium supply is small and strategically constrained.' severity: medium
• issue: 'Hydrogen is out of scope for a pure methalox Starship analysis, but not for an "Earth chemical rockets" ceiling. Hydrolox alternatives require LH2 production, liquefaction electricity, insulation, boiloff, and large-volume tankage.' severity: medium
• issue: 'Cryogenic insulation, vacuum-jacketed piping, perlite/MLI, tank farms, pumps, valves, and high-flow loading hardware are missing from the industrial-input list.' severity: medium
• issue: 'Avionics, IMUs, flight computers, sensors, batteries, harnessing, and space-rated electronics are omitted. Probably not binding near term, but relevant at 10^5-10^7 launches/yr.' severity: low
• issue: 'Copper and high-temperature alloys are omitted from the Raptor constraint. Combustion chambers, turbomachinery, wiring, power systems, and pad electrical infrastructure could become material at aircraft- or auto-scale engine production.' severity: medium
• issue: 'The analysis mixes US and global denominators: US LOX/grid/gas are used for some ceilings, global oxygen for others, and cosmic-scale ASU power is compared to US electricity despite being framed as an Earth-scale limit.' severity: medium
• issue: 'Ground tests, static fires, chilldown losses, boiloff, scrubbed launches, reserve propellant, and production rejects are excluded, so per-launch propellant and engine demand are optimistic.' severity: medium

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

leaf: q1-earth-industrial-ceiling pass: 05-consistency started: 2026-06-08T23:10:00Z ended: 2026-06-08T23:12:00Z researcher: claude-opus-4-7

Pass 05 — Consistency

Sibling-leaf claim-graph check. No-op at present: q1 is the first leaf in the lunar-throughput-multiplier report to reach this sub-pass. No other leaf has status: reviewed yet (q2, q3, q5, q6, q8a, q8b are still pending; the synthesis q9 depends on them all). There are no sibling claims.yaml files to compare against.

Expected cross-leaf interactions (to be checked when siblings reach reviewed)

When parallel-batch leaf passes complete, the cross-consistency pass should check:

• q1 ↔ q2: Is q1's atmospheric-non-constraint stance (q1 treats methane as supply-only, not pollution) consistent with q2's atmospheric ceiling? q1 says US NG saturates at 545k launches/yr; q2 should bind much earlier from ozone / reentry chemistry. Expected: q2 binds well before q1.
• q1 ↔ q3: q1 uses 100 t/launch (Block 3 reusable LEO) as the payload baseline. q3 will compute payload fractions per propulsion class × per destination. Cross-check: q1's LEO assumption × q3's destination-fraction = the actual t/yr to each destination.
• q1 ↔ q8a: If Moon supplies LOX to LEO depots, q1's LOX denominator shifts upward by the Moon's LOX export rate. Cross-check that q8a's LOX-export-to-LEO is consistent with q1's per-launch LOX demand.
• q1 ↔ q5: No direct dependency. q5 covers regolith mining ceiling; q1 covers Earth industrial inputs. Sibling, not chain.
• q1 ↔ q6: Indirect. q6's mass-driver throughput at maturity might draw on Earth-shipped vitamins (per q5), so q1's industrial-input ceilings might constrain how fast lunar industry can be bootstrapped. Worth checking at synthesis.

Action

No claims modified. leaf.yaml.contradictions_with[] remains empty. Re-run this sub-pass (or trigger cross-consistency) after the next leaf reaches reviewed.

leaf: q1-earth-industrial-ceiling pass: 03-reconcile started: 2026-06-08T22:30:00Z ended: 2026-06-08T22:55:00Z researcher: claude-opus-4-7

Pass 03 — Reconcile

Sources unsealed. Compare derived numbers from pass-02-calc.md to what extracts in sources/ say. Update claims.yaml.

Agreement table

Disagreements

No substantive disagreements found. All derived numbers reconcile with the source extracts within the documented Block-2/Block-3 sensitivity range and the explicit aspirational-vs-current overlay on the engine and pad parameters. The qualitative ordering (LOX binds first; engines second; pads third; methane / steel / ASU power not binding at the cadences analysed) is corroborated by every source that touches on it.

Notable cross-references

1. Handmer's "67,000 launches/yr for 1 TW orbital solar" is a useful independent triangulation. At 100 t/launch this is 6.7 Mt/yr to LEO, which sits comfortably below my derived 1-100 Mt/yr ceiling but above the LOX hard ceiling (~2,300 launches/yr from current US O₂). So serving 1 TW of orbital solar annually requires building dedicated LOX capacity (per the calc's "dedicated ASUs × 1,350" line for 100 Mt/yr scale). Handmer treats this as plausible.

2. Musk's 10,000 ships/yr is a production target not a flight target. At 100-flight reuse, 10,000 ships/yr of production supports 1,000,000 launches/yr in steady-state (after fleet replacement). My derived ceilings show LOX binds at ~2,300 launches/yr from current US O₂; reaching 10⁶ launches/yr requires 1,350 Baytown-class ASUs ($1.1T capex) plus ~2,700 pads. Musk's 10,000-ships target is bounded by LOX, not by ship production.

3. The SDC anchor's "2,660 launches/yr LOX hard ceiling" is from a Block-2 baseline (3,870 t LOX/launch); my Block-3 calc gives 2,330 launches/yr. Different Blocks; same OOM; consistent.

Resolution

• Hold all derived numbers and the qualitative ordering. No re-derivation needed.
• Promote Handmer's 67k-launches-for-1-TW-orbital-solar from "context" to "evidence" — it bounds the realistic plausible-demand-end of the question.
• Promote Musk's 10k-ships/yr as a public-figure target that the calc shows is LOX-bound, not production-bound. This is a load-bearing factual claim for tier-B source review.
• Flag for source-review pass: the EPA O2 PDF retrieval failure means our most-load-bearing US-LOX-supply number rests on a secondary citation. A re-pass should fetch the EPA HTML or use the alternative USGS / industrial-gas-association primary data.

claims.yaml — new entries

Adding 11 claims to claims.yaml. Each is derived from pass-02-calc.md and reconciled with at least one source from the research pass. All audit.gpt entries refer to pass-02-audit.md and pass-02-response.md.

Next sub-pass

--pass leaf --leaf q1-earth-industrial-ceiling --sub source-review — per-tier review with verdict taxonomy.

leaf: q1-earth-industrial-ceiling pass: 02-calc started: 2026-06-08T21:40:00Z ended: 2026-06-08T22:10:00Z researcher: claude-opus-4-7 sources_sealed: true

Pass 02 — Calc (sources sealed)

Methodology framing: This pass derives per-launch industrial-input demand from physics + vehicle spec (first-principles arithmetic on stoichiometry, propellant mass, engine count, vehicle dry mass). Per-input supply denominators (US/global O₂, US NG, US grid electricity, stainless steel, engine production target, pad count) are stated as explicit assumptions. The pass is "first-principles given explicit-stated denominators," not pure first-principles across the board. Sources sealed: no sources/ reading during this pass. Codex audit (pass-02-audit.md) → response (pass-02-response.md) → this revised calc.

Goal

For each industrial input — LOX, methane / natural gas, Raptor engines, launch pads, stainless steel, ASU electricity — derive the launches/yr (and equivalent kt/yr to LEO) at which the input binds at:

• Soft ceiling: 5% of relevant US/global supply (point at which the rocket industry causes detectable disruption in other industrial demand).
• Practical ceiling: 50% of relevant supply (point at which the rocket industry has reorganised the supply chain around itself).
• Hard ceiling: 100% of relevant supply (saturation; everything beyond requires building dedicated new capacity).

Then derive the ordering — which input binds first.

Assumptions (load-bearing)

Each assumption has an explicit physical or definitional basis. Sources sealed; these come from standard rocketry and engineering reasoning.

1. Reference vehicle: Block 3 Starship + Super Heavy stack with 100 t reusable payload to LEO. This is the design point that bounds the realistic mature ceiling; sensitivity to Block 2 (35 t payload, ~5,150 t propellant) is bounded by a factor of ~3 in propellant per kg of payload.
2. Per-stack propellant mass: approximately 5,650 metric tons (Block 3 baseline). Sensitivity: ±10% across announced Block 3 / late-Block 2 numbers.
3. LOX-to-LCH4 mass ratio: 3.6 : 1 (Raptor operating mixture ratio — fuel-rich relative to the 4.0:1 stoichiometric methalox ratio). LOX per stack: 5,650 × 3.6/4.6 = 4,422 t. LCH4 per stack: 5,650 × 1/4.6 = 1,228 t.
4. Natural gas → methane conversion: pipeline-grade natural gas in the US is 95%+ methane by volume; methane density at STP ≈ 0.7 kg/m³ → 1,430 m³/t methane → at 1 m³ ≈ 35.3 scf, per-stack natural-gas equivalent ≈ 1,228 t × 1,430 m³/t × 35.3 scf/m³ ÷ 10⁶ ≈ 62 million scf (matches industry analyses within rounding).
5. U.S. industrial oxygen production: approximately 10.3 Mt/year (most-recent comprehensive U.S. government figure). Sensitivity: ±10% across the supply-chain literature.
6. Global industrial oxygen production: approximately 90 Mt/year. Sensitivity: ±10%.
7. U.S. natural gas consumption: approximately 33.5 trillion cubic feet (Tcf) per year (most-recent EIA annual total). Sensitivity: ±5%.
8. Per-stack engine count: 39 Raptor engines (33 booster + 6 ship). Stable.
9. Per-stack dry steel mass: approximately 275 t (Block 2 reference). Sensitivity: ±20%.
10. U.S. stainless steel production: approximately 1.95 Mt/year. Sensitivity: ±5%.
11. Global stainless steel production: approximately 62.6 Mt/year. Sensitivity: ±5%.
12. Engine reuse-life baseline: 100 flights per engine (aspirational — demonstrated record on Raptor is well below this; Falcon 9 Merlin record sits at ~33 reuses per booster as of early 2026). Sensitivity: at 10-flight reuse the engine ceiling drops 10× (to ~256 launches/yr under current production target).
13. ASU electricity intensity: approximately 300 kWh per tonne of oxygen at modern large-cryogenic-distillation plants. Sensitivity: 150–800 kWh/t range across literature.
14. Pad cadence: 1 launch per pad per day (achievable-but-aspirational steady-state); peak aspirational is 3 per pad per day. Reference 50 pads globally as the maximum realistic pad-count (current operating launch sites worldwide is ~30; with major expansion 100 is the upper bound).
15. U.S. electricity generation: approximately 4,200 TWh per year.

Derivation

Step 1 — per-launch material consumption

# Block 3 Starship per-launch
propellant_total_t = 5650               # tonnes per stack
lox_ratio = 3.6 / 4.6                   # mass fraction of total propellant
lch4_ratio = 1.0 / 4.6
lox_per_launch_t = propellant_total_t * lox_ratio       # 4422 t
lch4_per_launch_t = propellant_total_t * lch4_ratio     # 1228 t

payload_per_launch_t = 100              # reusable Block 3
engines_per_launch_total = 39
engine_reuse_life = 100                 # flights per engine (assumption)
engines_per_launch_amortised = engines_per_launch_total / engine_reuse_life  # 0.39

stack_steel_t = 275
# For reusable operation, steel is amortised over engine_reuse_life as well
# (vehicle reuse ≈ engine reuse, both ~100 flights)
steel_per_launch_amortised_t = stack_steel_t / engine_reuse_life            # 2.75 t/launch

Step 2 — supply-side denominators

# US production (annual tonnes / tcf)
us_lox_supply_Mtyr = 10.3
world_lox_supply_Mtyr = 90.0
us_ng_consumption_Tcfyr = 33.5
us_stainless_supply_Mtyr = 1.95
world_stainless_supply_Mtyr = 62.6
us_electricity_supply_TWhyr = 4200

# Conversion: 1 t LCH4 ≈ 50,200 scf NG (95%+ methane assumption, 1.4 m^3/t × 1.42 m^3/scf)
# Mobius cross-check: 1,278 t methane = 62.24 mmscf → 48,700 scf/t. Use 50k as round.
scf_per_t_lch4 = 50000
us_ng_consumption_t_methane_equiv_yr = us_ng_consumption_Tcfyr * 1e12 / scf_per_t_lch4  # ~6.7e8 t/yr methane-equiv

# Plausible global production launch pad count
plausible_pad_count_max = 50      # mature mid-21st-century maximum, current ~30
pads_per_year_aspirational = 365         # 1 launch / pad / day
pads_per_year_peak = 1095                # 3 launches / pad / day (Musk aspirational)

Step 3 — per-input launches-per-year ceilings (5%, 50%, 100% of supply denominator)

For each input: compute the launch rate at which the input demand equals X% of supply.

def ceiling(demand_per_launch, supply_per_year, fraction):
    """Return launches/yr at which demand = fraction × supply."""
    return supply_per_year * fraction / demand_per_launch

# LOX
lox_5pct = ceiling(lox_per_launch_t, us_lox_supply_Mtyr * 1e6, 0.05)        # 116 launches/yr
lox_50pct = ceiling(lox_per_launch_t, us_lox_supply_Mtyr * 1e6, 0.50)       # 1,164 launches/yr
lox_100pct_us = ceiling(lox_per_launch_t, us_lox_supply_Mtyr * 1e6, 1.0)    # 2,329 launches/yr
lox_100pct_world = ceiling(lox_per_launch_t, world_lox_supply_Mtyr * 1e6, 1.0)  # 20,353 launches/yr

# LCH4 / natural gas
lch4_us_ng_fraction = lch4_per_launch_t / us_ng_consumption_t_methane_equiv_yr  # ~1.83e-6 per launch
lch4_5pct_us = 0.05 / lch4_us_ng_fraction      # 27,300 launches/yr
lch4_50pct_us = 0.50 / lch4_us_ng_fraction     # 273,000 launches/yr
lch4_100pct_us = 1.0 / lch4_us_ng_fraction      # 546,000 launches/yr

# Engines (production rate)
engine_production_target_per_year = 1000      # SpaceX stated production target (aspirational)
engines_amortised_per_launch = 39 / 100        # engines_per_stack / engine_reuse_life
# 5% / 50% / 100% of stated 1000/yr target:
engines_5pct  = 0.05 * 1000 / engines_amortised_per_launch   # 128 launches/yr
engines_50pct = 0.50 * 1000 / engines_amortised_per_launch   # 1,282 launches/yr
engines_100pct = 1.00 * 1000 / engines_amortised_per_launch  # 2,564 launches/yr
# Sensitivity at 10-flight reuse: ceilings drop 10× to 12.8 / 128 / 256 launches/yr
# At 10× production scale-up (10,000 engines/yr): ceilings rise 10× to 1,280 / 12,820 / 25,640

# Pads
# 5% / 50% / 100% of plausible mature global pad count (50 pads)
pads_5pct  = 0.05 * plausible_pad_count_max * pads_per_year_aspirational   # 912 launches/yr
pads_50pct = 0.50 * plausible_pad_count_max * pads_per_year_aspirational   # 9,125 launches/yr
pads_100pct = plausible_pad_count_max * pads_per_year_aspirational         # 18,250 launches/yr
# at 3 launches/pad/day (peak aspirational): 2,737 / 27,375 / 54,750

# Stainless steel (reusable, amortised over 100 flights)
steel_5pct_us  = ceiling(steel_per_launch_amortised_t, us_stainless_supply_Mtyr * 1e6, 0.05)  # 35,500 launches/yr
steel_50pct_us = ceiling(steel_per_launch_amortised_t, us_stainless_supply_Mtyr * 1e6, 0.50)  # 354,500 launches/yr
steel_100pct_us = ceiling(steel_per_launch_amortised_t, us_stainless_supply_Mtyr * 1e6, 1.0) # 709,000 launches/yr
# Stainless steel (expendable; full stack per launch)
steel_5pct_us_expendable = ceiling(stack_steel_t, us_stainless_supply_Mtyr * 1e6, 0.05)      # 355 launches/yr
steel_50pct_us_expendable = ceiling(stack_steel_t, us_stainless_supply_Mtyr * 1e6, 0.50)     # 3,545 launches/yr
steel_100pct_us_expendable = ceiling(stack_steel_t, us_stainless_supply_Mtyr * 1e6, 1.0)     # 7,090 launches/yr

# ASU electricity (grid power for cryogenic distillation)
asu_kwh_per_t_o2 = 300
asu_electricity_per_launch_kWh = asu_kwh_per_t_o2 * lox_per_launch_t            # 1,327,000 kWh = 1.33 GWh/launch
asu_5pct_us_grid = 0.05 * us_electricity_supply_TWhyr * 1e9 / asu_electricity_per_launch_kWh  # 158,300 launches/yr
asu_50pct_us_grid = 0.50 * us_electricity_supply_TWhyr * 1e9 / asu_electricity_per_launch_kWh  # 1,583,000 launches/yr
asu_100pct_us_grid = 1.0 * us_electricity_supply_TWhyr * 1e9 / asu_electricity_per_launch_kWh  # 3,165,000 launches/yr

Step 4 — verify with Python

python3 -c "
propellant_total_t = 5650; lox_ratio = 3.6/4.6; lch4_ratio = 1.0/4.6
lox_per_launch_t = propellant_total_t * lox_ratio
lch4_per_launch_t = propellant_total_t * lch4_ratio
us_lox = 10.3e6; world_lox = 90e6; us_ng_Tcf = 33.5
us_stainless = 1.95e6; us_electricity_TWh = 4200
scf_per_t = 50000
us_ng_t_methane = us_ng_Tcf * 1e12 / scf_per_t
asu_kwh = 300

print(f'LOX per launch (t): {lox_per_launch_t:.0f}')
print(f'LCH4 per launch (t): {lch4_per_launch_t:.0f}')
print(f'LOX 5%/50%/100% US ceiling (launches/yr): {us_lox*0.05/lox_per_launch_t:.0f} / {us_lox*0.50/lox_per_launch_t:.0f} / {us_lox/lox_per_launch_t:.0f}')
print(f'LOX 100% world ceiling: {world_lox/lox_per_launch_t:.0f}')
print(f'LCH4 5%/100% US-NG ceiling: {us_ng_t_methane*0.05/lch4_per_launch_t:.0f} / {us_ng_t_methane/lch4_per_launch_t:.0f}')
print(f'Engines at 1000/yr at 100-flight life: {1000/(39/100):.0f} launches/yr')
print(f'Pads (50 pads × 1/day): {50*365} launches/yr; at 3/day: {50*1095} launches/yr')
print(f'Steel reusable 100% US (275 t/100 flights): {us_stainless/(275/100):.0f} launches/yr')
print(f'Steel expendable 100% US: {us_stainless/275:.0f} launches/yr')
print(f'ASU electricity per launch (GWh): {asu_kwh*lox_per_launch_t/1e6:.2f}')
print(f'ASU 5%/100% US grid: {0.05*us_electricity_TWh*1e9/(asu_kwh*lox_per_launch_t):.0f} / {us_electricity_TWh*1e9/(asu_kwh*lox_per_launch_t):.0f} launches/yr')
"

Output:

LOX per launch (t): 4422
LCH4 per launch (t): 1228
LOX 5%/50%/100% US ceiling (launches/yr): 116 / 1164 / 2329
LOX 100% world ceiling: 20353
LCH4 5%/100% US-NG ceiling: 27282 / 545635
Engines at 1000/yr at 100-flight life: 2564 launches/yr
Pads (50 pads × 1/day): 18250 launches/yr; at 3/day: 54750 launches/yr
Steel reusable 100% US (275 t/100 flights): 709091 launches/yr
Steel expendable 100% US: 7091 launches/yr
ASU electricity per launch (GWh): 1.33
ASU 5%/100% US grid: 158296 / 3165910 launches/yr

Step 5 — equivalent kt/yr LEO at each ceiling

# Multiply launches/yr × 100 t payload (Block 3 reusable)
# Conversion: 1 launch = 100 t LEO; 1,000 launches = 100 kt; 10,000 = 1 Mt

Result — binding-input ordering

LOX binds first, by a wide margin. Soft ceiling at ~120 launches/yr (≈ 12 kt LEO/yr); hard ceiling at ~2,300 launches/yr (≈ 230 kt LEO/yr) without new dedicated ASU capacity. Building new ASU capacity adds about 0.7-2.0 Mt/yr LOX per Air-Liquide-Baytown-class facility, supporting an additional ~150-450 launches/yr per new plant.

Engine production binds next at ~2,600 launches/yr under the 100-flight-engine-life assumption and current ~1,000 engines/yr target. Aircraft-industry-scale production (10⁴-10⁵ engines/yr) raises this ceiling proportionally, but the supply chain for 300-bar full-flow staged combustion engines does not have any current analog.

Pad cadence is the structural ceiling at ~10⁴-10⁵ launches/yr, even with aspirational 1-3 launches/day per pad and 50-100 pads globally. Beyond ~50,000 launches/yr you are building dozens of new pad complexes per year.

Methane / natural gas is not the binding constraint under any realistic scenario. Even saturation (100% of current US natural gas going to Starship) supports 545,000 launches/yr — about 50 Mt/yr to LEO.

Steel is binding only for expendable operations (~7,100 launches/yr saturates US stainless). For reusable operations (vehicle life ≈ engine life ≈ 100 flights), steel is non-binding until ~700,000 launches/yr.

ASU electricity is non-binding until ~3 million launches/yr (saturation of current US grid). Long before that, the LOX-supply constraint has forced new ASUs to be built with dedicated generation co-located.

Hard physical ceiling — per scale, what each input must look like

Cleaner numbers than the original draft (which conflated 10⁸ t/yr and 10⁹ t/yr scales). Per scale, at 100 t/launch and 100-flight reuse:

The Earth chemical-rocket throughput ceiling sits at roughly 1-100 Mt/yr (10⁶-10⁸ tonnes/year) to LEO with civilizational-scale industrial mobilization. The high end (100 Mt/yr) requires: dedicated LOX industry ~15× current global O₂ capacity (4,420 Mt/yr vs ~90 Mt/yr today), ~1,350 Baytown-class ASUs, ~2,740 launch pads (vs ~30 active sites today), ~390,000 engines/year, and ~130 GW of dedicated ASU power (~3% of current US grid). Each of these is a multiple-OOM scale-up from anything currently existing.

Result — cosmic-scale verdict

At cosmic-scale (Tt/yr to LEO = 10⁹ t/yr):

• LEO mass: 10⁹ t/yr at 100 t/launch = 10⁷ launches/yr.
• LOX demand: 4.42 × 10¹⁰ t/yr = ~491× current global O2 production.
• ASU power: ~1.33 × 10¹³ kWh/yr = ~13,300 TWh/yr ≈ ~3.2× current US electricity generation.
• Pad infrastructure: 10⁷ launches/yr ÷ 1,000 launches/pad-yr ≈ 10,000 pads — beyond any plausible Earth pad-site map (which has historically held ~30 active sites).

Earth chemical rockets cannot plausibly deliver Tt/yr to LEO. The ceiling sits 1-3 orders of magnitude below Tt/yr (Tt/yr = 10⁹ t/yr; chemical ceiling 10⁶-10⁸ t/yr).

The categorical-infeasibility framing rests on what reaching even the 10⁸ t/yr ceiling already requires (dedicated LOX industry ~15× current global, ~2,700 pads, ~400,000 engines/yr), and what extending another 1-3 OOM beyond that would entail: a LOX industry 50-500× current global O₂; ~10⁴ launch pads; ~10⁷ engines/yr; ~1 TW dedicated ASU power. These exceed humanity's largest current industrial systems (steel ~2 Gt/yr global, oil ~5 Gt/yr global, electricity ~30,000 TWh/yr global) by orders of magnitude.

This is the load-bearing answer for the q9 synthesis: Tt/yr cosmic-scale demand for off-Earth mass cannot be served by Earth chemical-rocket launch. The Moon (or another non-Earth-chemical channel) is necessary, not merely cheaper.

Confidence

• High confidence in the binding-input ordering (LOX first, then engines, then pads). This rests on stoichiometry, public industrial-gas data, and pad-cadence arithmetic — all robust to ±20% assumption variation.
• Medium confidence in the specific kt/yr numerical ceilings, because (a) Block 3 specs are still being finalized, (b) ASU energy intensity has a 5× literature range, (c) future industrial scaling is path-dependent.
• High confidence in the cosmic-scale verdict: Earth chemical cannot deliver Tt/yr to LEO under any plausible scaling.

Caveats and known omissions (Codex audit informed)

The following constraints were excluded from this calc and are not the binding constraint at the scales analysed, but should be visible to the reader:

• Merchant LOX vs bulk industrial O₂: The 10.3 Mt/yr US figure is total US oxygen production. Merchant liquid oxygen capacity (storage, transport, rail/road loading, on-site cryogenic distillation hours) is a subset of that, and the rocket-grade LOX deliverable rate is lower still. The LOX ceiling quoted is an upper bound; the operational rocket-LOX bottleneck is somewhat tighter.
• Methane liquefaction vs pipeline NG: US natural gas supply is enormous, but rocket-grade liquefied CH₄ requires dedicated liquefaction capacity, low-temperature storage, and methane-grade purification beyond pipeline standards. None of these is individually binding before LOX in the scales analysed.
• Helium supply: Starship is autogenous-pressurized in flight but the launch supply chain requires helium for purge, leak testing, GSE pneumatics, and ground systems. Helium supply is small and strategically constrained globally (~30 kt/yr); the per-launch helium load isn't computed here but is a known omission for the audit-pass to-do.
• Cryogenic insulation, vacuum-jacketed plumbing, MLI: Tank farms, propellant transfer lines, and GSE require non-trivial cryogenic insulation infrastructure. Not bounded here.
• Avionics, IMUs, flight computers, sensors, batteries, harnessing: Not binding at the cadence-scales analysed but worth flagging at 10⁵+ launches/yr.
• Copper and high-temperature alloys: Combustion chambers, turbomachinery wiring, ground-power systems. Material at aircraft-industry-scale engine production.
• US vs global denominators: This calc uses US figures (LOX, NG, electricity, stainless) as the first-bottleneck reference because SpaceX is US-based and US bottlenecks bind first. Global figures appear only for the world-LOX line and the cosmic-scale comparison. A more thorough audit pass would use a country-by-country supply matrix.
• Ground-system overhead: Per-launch propellant demand here is the vehicle load. Real ground demand is ~1.2-1.5× higher because of static fires, chilldown, boiloff, scrubs, reserves, and rejects. The ceiling numbers therefore overstate the true rocket-launch ceiling by ~20-50%.
• Hydrolox alternative: Out of scope (the report is calibrated to methalox baseline per Avi's framing); the LOX-binding result is broadly invariant since LH₂ also burns with LOX at ~6-8:1 mass ratio.

Derived claims (to add to claims.yaml in pass-03-reconcile)

• q1.c1 [derived, high]: LOX binds first; soft ceiling ~120 launches/yr from current US O2 supply, hard ceiling ~2,300 launches/yr.
• q1.c2 [derived, high]: Engine production binds next at ~2,600 launches/yr under 100-flight engine reuse and current 1,000-engine/yr target.
• q1.c3 [derived, high]: Pad cadence ceiling is ~10⁴-10⁵ launches/yr (50 pads × 1-3 launches/day).
• q1.c4 [derived, high]: Methane/NG is not the binding constraint at any realistic Starship cadence (US NG saturates at ~550,000 launches/yr).
• q1.c5 [derived, high]: Stainless steel is binding only for expendable operations; non-binding for reusable.
• q1.c6 [derived, medium]: ASU electricity demand reaches 5% of US grid at ~160,000 launches/yr; not a near-term constraint.
• q1.c7 [derived, high]: With civilizational-scale industrial mobilization (dedicated ASUs ~50× global, 10⁴-10⁵ pads, 10⁷ engines/yr, ~150 GW dedicated ASU power), Earth chemical-rocket throughput tops out at ~10⁶-10⁸ t/yr to LEO.
• q1.c8 [derived, high]: Earth chemical rockets cannot deliver Tt/yr (10⁹ t/yr) to LEO under any plausible industrial scaling. 4-5 OOM below cosmic-scale demand.

Next sub-pass

--pass leaf --leaf q1-earth-industrial-ceiling --sub reconcile — sources now unsealed; compare these derived numbers to what sources say.

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

overall: verdict: weak summary: 'Near-term arithmetic is mostly internally consistent, but the pass is not genuinely first-principles and the high-scale ceiling claims contain major inconsistencies. The LOX-first ordering is plausible under the stated assumptions, while the 10^6-10^8 t/yr and Tt/yr conclusions are under-derived.'

findings:

• target: 'method.first_principles' target_kind: method verdict: fail reason: 'Many key denominators are borrowed, not derived: US/global oxygen production, US gas consumption, stainless production, electricity generation, Raptor production target, pad count, and Baytown-class ASU capacity. That is acceptable as assumptions, but contradicts the pass framing.' severity: high

• target: 'goal.5_50_100_ceiling_coverage' target_kind: completeness verdict: fail reason: 'The stated goal requires 5%, 50%, and 100% ceilings for each input. Several inputs omit 50% cases, and engines/pads are not expressed as fractions of a defined supply denominator.' severity: high

• target: 'q1.c1' target_kind: claim verdict: partial reason: 'The LOX arithmetic is correct given 5,650 t propellant, O/F 3.6, and 10.3 Mt/yr oxygen supply. But the denominator is industrial oxygen, not necessarily merchant liquid oxygen capacity, storage, transport, or load-rate capacity, so the claimed LOX ceiling may be overstated.' severity: medium

• target: 'assumption.3' target_kind: assumption verdict: partial reason: 'The 3.6:1 methalox mixture ratio is plausible for Raptor-like engines, but it is not stoichiometric; stoichiometric CH4/O2 is 4.0:1 by mass. The numerical impact is small, but the stated physical basis is wrong.' severity: low

• target: 'step2.ng_conversion_comment' target_kind: calculation verdict: partial reason: 'The computed 50,000 scf/t methane is approximately right, but the comment gives garbled units: it should be about 1,400 m3 gas per tonne methane divided by 0.0283 m3/scf, not 1.4 m3/t times 1.42 m3/scf.' severity: low

• target: 'q1.c2' target_kind: claim verdict: weak reason: 'The 2,564 launches/yr calculation follows from 1,000 engines/yr and 39 engines amortized over 100 flights. But both the 1,000-engine target and 100-flight life are borrowed/aspirational, and engine bill-of-material constraints are not analyzed.' severity: medium

• target: 'q1.c3' target_kind: claim verdict: partial reason: 'The pad arithmetic is correct for 50-100 pads at 1-3 launches/day. The claim is still assumption-driven and omits pad refurbishment, deluge water, GSE, local range constraints, methane/LOX loading infrastructure, and site geography.' severity: medium

• target: 'q1.c4' target_kind: claim verdict: partial reason: 'Natural-gas feedstock is correctly shown as non-binding relative to LOX under the stated assumptions. However, rocket-grade LCH4 purification, liquefaction capacity, storage, boiloff, and liquefaction electricity are not included.' severity: medium

• target: 'q1.c5' target_kind: claim verdict: partial reason: 'The reusable steel arithmetic is internally correct if a 275 t stack lasts 100 flights. It omits fleet buildout, attrition, scrap/yield, specific stainless alloy availability, and steel in pads, tanks, ASUs, and GSE.' severity: low

• target: 'q1.c6' target_kind: claim verdict: supports reason: 'The ASU electricity calculation is dimensionally correct: 300 kWh/t times 4,422 t LOX gives about 1.33 GWh/launch, making grid electricity non-binding before LOX capacity. Caveat: this excludes methane liquefaction and some liquid-product/storage losses.' severity: low

• target: 'q1.c7' target_kind: claim verdict: contradicted reason: 'The high-end 100 Mt/yr case equals about 1,000,000 launches/yr. At 100-flight reuse that needs about 390,000 engines/yr, not 10^7, and at 1,000 launches/pad-yr it needs about 1,000 pads, not 10^5.' severity: high

• target: 'q1.c8' target_kind: claim verdict: weak reason: 'The Tt/yr arithmetic is mostly correct, but the "4-5 OOM below" statement contradicts the stated 10^6-10^8 t/yr ceiling, which is only 1-3 OOM below 10^9 t/yr. The impossibility conclusion is plausible but not established by the presented derivation.' severity: high

notes:

• issue: 'Helium is omitted. Starship may use autogenous pressurization, but purge, testing, leak-checking, pneumatic, and ground-system helium demand should be explicitly bounded because helium supply is small and strategically constrained.' severity: medium
• issue: 'Hydrogen is out of scope for a pure methalox Starship analysis, but not for an "Earth chemical rockets" ceiling. Hydrolox alternatives require LH2 production, liquefaction electricity, insulation, boiloff, and large-volume tankage.' severity: medium
• issue: 'Cryogenic insulation, vacuum-jacketed piping, perlite/MLI, tank farms, pumps, valves, and high-flow loading hardware are missing from the industrial-input list.' severity: medium
• issue: 'Avionics, IMUs, flight computers, sensors, batteries, harnessing, and space-rated electronics are omitted. Probably not binding near term, but relevant at 10^5-10^7 launches/yr.' severity: low
• issue: 'Copper and high-temperature alloys are omitted from the Raptor constraint. Combustion chambers, turbomachinery, wiring, power systems, and pad electrical infrastructure could become material at aircraft- or auto-scale engine production.' severity: medium
• issue: 'The analysis mixes US and global denominators: US LOX/grid/gas are used for some ceilings, global oxygen for others, and cosmic-scale ASU power is compared to US electricity despite being framed as an Earth-scale limit.' severity: medium
• issue: 'Ground tests, static fires, chilldown losses, boiloff, scrubbed launches, reserve propellant, and production rejects are excluded, so per-launch propellant and engine demand are optimistic.' severity: medium

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

leaf: q1-earth-industrial-ceiling pass: 02-response responding_to: pass-02-audit.md auditor: codex / gpt-5.5 responder: claude-opus-4-7

Pass 02 — Response to Codex audit

Accepted (will update calc)

• method.first_principles (fail, high): The pass is "first-principles given explicit-stated supply denominators," not pure first-principles. The denominators (US/global O₂, US NG, stainless, grid electricity, engine production target, pad count) are stated assumptions, not derived. The frame should be: per-launch demand is derived from physics + vehicle spec; per-input supply is observed at current state. Will reframe the calc's intro and Assumptions block.

• goal.5_50_100_ceiling_coverage (fail, high): I omitted the 50% ceiling for engines, pads, steel, and ASU power. Will add to the table.

• q1.c1 (partial, medium): "Industrial O₂ production" is a coarser denominator than merchant liquid oxygen capacity (storage, transport, road/rail load-rate, on-site distillation column hours). The actual rocket-relevant LOX ceiling is somewhat lower than the headline number. Will add a caveat-paragraph: bulk-O₂ supply is the upper bound on rocket-grade-LOX supply; the merchant-LOX-loadable bottleneck is unmeasured here.

• assumption.3 (partial, low): Stoichiometric CH₄/O₂ is 4.0:1 by mass; 3.6:1 is slightly fuel-rich (sub-stoichiometric), not "slightly oxidiser-rich." Numerical impact is small but the physical-basis statement was wrong. Will fix.

• step2.ng_conversion_comment (partial, low): Comment garbled the m³/scf conversion. Density of methane at STP ≈ 0.7 kg/m³ → 1.43 m³/kg → 1,430 m³/t methane; 1 m³ = 35.3 scf, so 1,430 × 35.3 ≈ 50,500 scf/t. The numerical answer is right; the inline comment was wrong. Will rewrite the comment.

• q1.c2 (weak, medium): The 1,000 engines/yr target AND the 100-flight reuse life are aspirational, not measured. I assumed both in the calc but did not flag them as load-bearing. Will mark both explicitly and note: a more conservative 10-flight reuse drops the engine ceiling 10× to ~256 launches/yr.

• q1.c3 (partial, medium): Pad arithmetic omits deluge water, GSE, range constraints, propellant loading infrastructure, geography. Will caveat: pad count is a first-cut; the load-bearing constraint at high pad counts is global suitable-site availability + loading throughput, both probably ≥2× more binding than launch slot scheduling alone.

• q1.c4 (partial, medium): Bulk NG vs liquefied CH₄ at launch site are different bottlenecks. Will add: pipeline-to-LNG liquefaction capacity, on-site storage / boil-off, and methane-grade purity (rocket-grade vs. pipeline) are sub-constraints, none individually binding before LOX but worth noting.

• q1.c5 (partial, low): Steel for pads, tanks, ASUs, GSE is non-trivial and excluded. Will note briefly.

• q1.c7 (contradicted, high): This is the most material correction. At 10⁸ t/yr to LEO = 10⁶ launches/yr the right numbers are: ~390,000 engines/yr (not 10⁷), ~1,000 pads (not 10⁵), and the LOX ASU power is ~13 GW continuous (not 150 GW). My text conflated the 10⁸ t/yr scale with the cosmic 10⁹ t/yr scale. Will write a clean tier-by-tier table showing what each scale requires in each input.

• q1.c8 (weak, high): "4-5 OOM below" is wrong arithmetic. Tt/yr = 10⁹ t/yr; Earth chemical ceiling 10⁶-10⁸ t/yr; difference is 1-3 OOM, not 4-5. The qualitative impossibility conclusion still stands because (a) reaching even the 10⁸ t/yr ceiling requires civilizational-scale rebuilding, and (b) extending another 1-3 OOM means dedicated LOX industry ~50× current global O₂, ~10⁴ pads, ~10⁷ engines/yr — categorically larger than humanity's largest current industrial undertakings (steel, oil refining, electricity). Will rewrite the conclusion with the correct OOM count and the right categorical-infeasibility framing.

• notes — Helium (medium): Real omission. Starship is autogenous-pressurized at flight, but helium is needed for purge, GSE, leak testing, pneumatics. Helium supply is small and strategically constrained globally. Will flag as a known-omission in the calc footer and as a candidate for the source-review or reconcile pass.

• notes — Cryogenic insulation, GSE, plumbing (medium): Real omission. Will list as known-omissions footer.

• notes — Avionics, IMUs, batteries (low): Real omission. Probably not binding ≤10⁵ launches/yr. Will footer-flag.

• notes — Copper, high-temperature alloys (medium): Real for engine bills-of-material at aircraft-scale production. Will footer-flag.

• notes — US vs global denominator mixing (medium): I used US LOX, US grid, US NG for some ceilings and global O₂ for one. Will explicitly disclose the choice (US is the first bottleneck because SpaceX is US-based; global is the eventual bound).

• notes — scrubs/boiloff/static-fires/reserves (medium): Real overhead. Per-launch propellant demand quoted (4,422 t LOX) is the rocket vehicle's load, not the full ground-system draw. Realistic ground demand is ~1.2-1.5× higher. Will add a "ground-system multiplier" note.

Disputed / clarified

• notes — Hydrogen / hydrolox alternatives (medium): Out-of-scope by Avi's framing: the report is calibrated to chemical methalox (Starship-like) as the primary regime; ion / NTR / NPP belong to q3, hydrolox is not a baseline. A hydrolox sensitivity could swap CH₄ for LH₂ and the LOX number stays similar (slightly higher with stoichiometric H₂/O₂ at 1:8), but LH₂ production / liquefaction is a different supply chain not modelled here. Clarification: the calc baseline stays methalox; a future re-pass can add an LH₂-alternative sensitivity if relevant for the q9 synthesis.

What changes substantively

• Two high-severity findings (q1.c7, q1.c8) require visible corrections to the calc's "hard physical ceiling" paragraph and the cosmic-scale-verdict paragraph.
• One high-severity finding (method.first_principles) requires reframing the calc's intro.
• One high-severity finding (5/50/100 coverage) requires extending the per-input ceiling table.
• The rest are caveat-paragraph additions and footer-list-of-known-omissions.

What does not change

• The qualitative ordering — LOX binds first, then engines, then pads — is robust under audit. Codex did not contradict the ordering.
• The cosmic-scale verdict (Earth chemical cannot deliver Tt/yr to LEO) is robust. Codex flagged the OOM-count phrasing but not the conclusion.
• The Block 3 reference vehicle, the methalox stoichiometry, the per-launch arithmetic, the US/global supply denominators — all stand under audit modulo the caveats above.

Next action

Update pass-02-calc.md with the substantive fixes flagged here. Then proceed to pass-03-reconcile (sources unsealed; compare derived numbers to extracts).

leaf: q1-earth-industrial-ceiling pass: 01-research started: 2026-06-08T21:00:00Z ended: 2026-06-08T21:35:00Z researcher: claude-opus-4-7

Pass 01 — Research

Source-gather only. No claims written to claims.yaml. Calc and reconcile sub-passes come next.

Tier-grouped source list

Tier S — peer-reviewed primary / government primary

Tier A — peer-reviewed reviews / conference proceedings / credentialed preprints

None retrieved in this gather. Note: q1's primary substrate is government and industry data; academic peer-reviewed work on this exact topic (rocket-propellant industrial scaling) is sparse. The research pass should re-run with a wider academic search for Acta Astronautica, Journal of Spacecraft and Rockets, AIAA SciTech papers on launch-vehicle production-rate scaling.

Tier B — public figures (quote review)

Tier C — industry trade press / expert blogs / corporate technical

Tier D — mainstream press / non-public-figure

Tier E — orientation only (not reviewed; not citable as evidence)

• spacex-raptor-wiki — Wikipedia Raptor entry. Used to orient on engine specs and production targets. Will not appear as evidence in claims.yaml; primary press releases inside are the citable substrates and will be re-fetched in a future research re-pass if needed.

Topics surfaced (will feed topics.yaml at report level)

• lox-supply-bottleneck — US oxygen production capacity, ASU capex and energy intensity, scaling math
• methane-supply-bottleneck — US natural gas consumption baseline, per-launch demand, LNG liquefaction
• engine-production-rate — Raptor production rate, reuse-life implications
• pad-cadence-faa — pad turnaround, FAA-approved cadence at Boca Chica and KSC
• airframe-feedstock — stainless steel needs for reusable vs. expendable fleet
• asu-capacity-power — ASU electricity demand at scale, dedicated capacity vs shared
• aspirational-cadence-target — Musk, Handmer, industry-stated cadence targets to bound the question

Coverage assessment

Tier S coverage is thin. The two primary government data sources (EIA NG consumption, EPA O2 supply chain) are the load-bearing tier-S anchors. The EPA PDF failed to fetch cleanly — needs a retry with a different rendering path (HTML, or alternative U.S. industrial-gas data source like the U.S. Geological Survey Mineral Commodity Summaries, which sometimes covers industrial gases).

Tier A coverage is empty. A targeted academic search did not surface peer-reviewed work on the exact topic of launch-vehicle production-rate constraints in chemical industry; the literature instead lives in policy and trade-press venues. This is a finding in itself — q1's substrate is institutional and industrial data, not academic. Flag for the audit pass.

Tier C is the analytical backbone — Mobius, Thunder Said Energy, Handmer, plus our internal SDC compilation. These give us the per-launch numerator and per-input denominator math for the calc sub-pass.

Tier B is small but pointed — Musk's 10,000-ships target and Handmer's 10⁶ t/yr LEO trajectory are the two aspirational anchors against which the calc-pass ceiling estimates should be compared.

Coverage gaps to address in a future research re-pass:

1. Direct fetch of the EPA Oxygen Supply Chain Profile (HTML rendering of the report's summary tables, or USGS alternative).
2. Methane liquefaction capacity at Starbase / Boca Chica — Rio Grande LNG, Brownsville LNG; relevant for fast-cycle Starship operations.
3. Air separation literature on energy efficiency of next-generation cryogenic plants (membrane separation, oxygen ionic transport membrane technology) — could matter at 10⁴+ launches/yr scale.
4. Pad infrastructure scaling literature — Apollo-era studies on multiple-pad launch operations, ESA / Roscosmos data on pad turnaround.

Sub-pass deliverables

• 10 extract.md files in sources/ covering tier S/B/C/D + 1 tier E for orientation.
• 1 pass-01-research.md (this file).
• 0 entries in claims.yaml — claims come from calc + reconcile, not research.

Anti-pattern hygiene check (research-pass-specific)

• No claim-bearing prose in this pass. ✓ Confirmed: this file is pure source-gather summary.
• No source borrowing across passes yet. ✓ Confirmed: calc-pass sources will be sealed.
• All extracts have proper frontmatter (tier, type, peer_reviewed, venue, year, topics). ✓ Confirmed across 10 extracts.
• No hallucinated sources. ✓ Confirmed: every extract corresponds to a WebFetch result or workspace anchor. The two failed fetches (EPA PDF, Spacenews, Ars Technica) are explicitly flagged.

Next sub-pass

--pass leaf --leaf q1-earth-industrial-ceiling --sub calc — derive ceiling from first principles, sources sealed.