Experimental — Dyno-Sim is under active development. This is a research-grade run: a reduced-form model with disclosed dials, reporting relative, structural results — here the difference between two runs, not any single level, year, or outcome. Interrogate it as decision-support. Illustrative / decision-support, not a forecast. Figures below come from Comuvia's Dyno-Sim using its shipped, stylized sample scenarios — they illustrate the mechanism and that the tool works, not a forecast of any economy or any budget. Every number is a difference between two model runs. A client engagement calibrates the model to your data and question. This is decision support, not investment advice.

Updated 2026-08-13. The two-run differences were refreshed at a converged integration step (dt 0.0625), traceable to a committed converged run. (The inline charts and the explainer video are queued for regeneration at the converged step.)

Rearmament is back as a budget line across the West, and every headline arrives with an economic opinion bolted on: it will crowd out investment, or it will reindustrialize the country, or it will bankrupt the state. If you sit on an investment committee, a policy desk, or a corporate strategy team, none of those slogans tells you the one thing you need — the magnitude of the trade-off, and how much of it is real versus assumed.

Averages and op-eds can't answer that. A mechanism you can interrogate, stress, and reproduce can. So Comuvia's system doesn't write an essay about "guns versus butter": it runs a simulator Comuvia owns, and it shows the work below. This page was produced by the same AI-managed system that ran the simulation — we run the stack we advise on.

Concept diagram — the same 2026 recession hits both worlds; one keeps the budget flat ("Do Nothing"), the other adds a +1.5%-of-GDP deficit-financed defense surge ("Rearmament"), so every downstream difference is caused by the surge alone.

The machine

The run uses Dyno-Sim's fiscal_monetary_v1 runtime — a stock-flow-consistent policy laboratory that composes a Treasury (counter-cyclical spending off a baseline civil path, an explicit tax share, and a budget constraint that issues the balance into debt) with a central bank (a Taylor-style rule, an effective lower bound, and asset purchases). Every quarter, the model checks its own accounting identities: spending, taxes, deficits and debt issuance have to balance, or the step is rejected. The capsule's own fidelity label travels with it — schematic — quoted, not softened.

Schematic of the fiscal_monetary_v1 lab — a Treasury and a central bank wired together over a shared ledger whose accounting identities must balance every quarter, or the step is rejected.

The experiment

The experiment is a controlled pair, not a prediction. Both worlds start from the same US-2025 configuration and are hit by the same mid-2026 recession shock. They differ in exactly one thing:

  • Baseline. The recession runs its course under the standard policy reaction functions.
  • Surge. The same world, plus a sustained +1.5 percentage-points-of-GDP, deficit-financed rearmament held through the recovery.

The committed SIPRI-via-World-Bank series grounds the scale, so the size of the surge isn't a number pulled from the air: 2024 US military spending is 3.42% of GDP; the surge lifts it toward ~4.9% — a Reagan-scale buildup, and still well below the same series' Cold-War peak of 9.4%.

US military spending as a share of GDP, 1960–2024 — today about 3.4%; the scenario adds +1.5 points to about 4.9% (Reagan-scale), still far below the 9.4% Cold-War peak.

What happens

Holding the recession fixed and toggling only the defense surge, the model reports the gap at the end of the horizon:

2030Q4 — surge world vs baseline (same recession)Difference
Unemployment2.1 points lower
Output2.7% higher
Public debt-to-GDP1.4 points higher

Read it as the trade-off in three numbers: the deficit-financed surge buys materially lower unemployment and higher output through the downturn, and the public-debt cost of doing so is small — because output growth absorbs most of the borrowing. These are differences between two model runs under one set of assumptions, not a forecast of what any government's spending will do.

Unemployment rate, 2027–2030 — both paths start near 10% after the 2026 recession; the defense-surge run ends about 2.1 points lower than doing nothing.

Output index (2025 = 100), 2027–2030 — the defense-surge run ends about 2.7% higher than the do-nothing baseline as the spending circulates beyond defense.

Public-debt ratio (% of GDP), 2027–2030 — the cost of the surge is about 1.4 points of GDP more debt by 2030, as both worlds deleverage into the recovery.

The trade-off is a dial, not a verdict

The debt cost is small because of one assumption: a fiscal multiplier of 0.55 — a scenario dial, explicitly not an estimate. That single parameter governs how much of each borrowed dollar shows up as output rather than debt, and the honest way to present a result that hinges on it is to show the sensitivity rather than bury it. Turn the multiplier down and the same surge buys less output and leaves more debt; turn it up and the trade-off flatters the surge further. The lab's job isn't to pick the "right" multiplier — it's to make the reader's choice of multiplier legible and its consequences computed.

The assumed fiscal multiplier shown as a disclosed dial set to ×0.55 — turn it down and the debt cost grows; the result is only as honest as the dial, so the dial is shown rather than buried.

How a forecast becomes a budget

Here is the part that scales beyond this one scenario. A military-spending forecast can be compiled straight into this laboratory — not pasted into a slide, but propagated through the model's identities. The milex bridge takes a spending trajectory (peak +0.53pp of GDP in the demo path), splits it into the budget, and finances it — with unit, budget-split, and financing identities validation-gated at machine precision (≤1e-12). The forecast becomes a fully accounted alternative world the model can run, not a headline number floating free of the balance sheet. (The model proposes; the accounting identity — the verifier — decides whether the alternative even balances.)

The accounting gate — spending = taxes + new borrowing + central-bank support, checked every quarter to a residual of about 2×10⁻¹⁶, far tighter than the 1×10⁻¹² tolerance that would otherwise reject the step.

Anchoring against real recessions

A schematic model earns trust by saying where it's weak. Anchored against the last two real US recessions, the honest reading cuts both ways: the lab's shock falls COVID-fast and cuts deeper than 2008 at onset — but without the surge, the modeled economy recovers slower than either real episode. That gap is a property of a schematic reaction function, and it's said out loud rather than hidden, because a model that only flatters itself is a brochure.

What it does not claim

The most important missing factors are registered, not buried:

  • demographics and entitlements,
  • the external sector (trade, the exchange rate),
  • an inflation-expectations channel,
  • the tax-versus-deficit financing experiment (this run finances the surge entirely with debt),
  • energy pass-through.

Each is a known absence with a place in the backlog, not a silent gap. Naming what the run does not model is how you know the rest is real — modeled, measured, composed, honest.

What this looks like pointed at your question

The run above answers a public question with a stylized scenario. A Simulation Decision Pack points the same machinery at a specific one — a defense-exposed portfolio, a supplier's demand outlook, a sovereign's fiscal-space question — with your multiplier assumptions made explicit and the trade-off computed rather than asserted. And when the modeled scenarios feed a larger choice, a Decision Map Workshop turns them into a decision map with a recommended path. Start from simulation-backed decision support — a run you can audit, not an opinion you have to trust.

Bottom line. Achieved: a stock-flow-consistent lab runs one recession twice and reports the rearmament trade-off as differences between two runs — unemployment −2.1pp, output +2.7%, public debt +1.4pp of GDP at 2030Q4 (converged dt=0.0625) — with a military-spending forecast compiled into the model through identities gated at machine precision. Business value: the guns-versus-butter question made quantitative, scenario-conditional and auditable, instead of a headline with an opinion attached.


Comuvia builds decision support on owned, accounting-consistent simulators. This piece uses a stylized sample scenario to show the method; engagements calibrate to your data. Every number is a difference between two runs, not a prediction. Not investment advice.