The score answers one question: how close is the U.S. financial system to a point where policymakers are forced to act?
That's a different question than recession probability, VIX, inflation, or Treasury yields taken in isolation. It's closer to proximity × severity × probability of policy-forcing financial-system stress — a judgment call that requires synthesizing Treasury-market mechanics, funding-market plumbing, and how the Fed and Treasury are actually behaving, not just what a single indicator says.
Each evaluation runs through three AI stages, each a different model, so no single model's judgment silently becomes the answer:
The score runs 0–100: 0–15 accommodative, 16–35 mild watchfulness, 36–55 real but contained tension, 56–75 persistent restrictive policy (active QT at multi-year-high yields with a genuinely undecided upcoming Fed decision already belongs here, even with no crisis signal — restrictive policy plus real two-sided uncertainty isn't calm), 76–90 policy-forcing danger zone, 91–100 market dysfunction or intervention underway. A second, related figure — how much Treasury/Fed policy has actually begun supplying liquidity — is tracked alongside it, since high stress with a rising policy response reads very differently than high stress with no response at all.
Alongside the headline score, each evaluation tracks three metrics explicitly against the last published one — not reasserted fresh every cycle, and each shown as a percentage: Controlled Break Risk (the odds current stress breaks into an uncontrolled crisis rather than staying policy-managed), Liquidity Response Probability (the odds the Fed/Treasury actually supplies new support soon — distinct from how much support is already underway), and Phase (1 = tightening/policy-threat, 2 = liquidity response beginning, 3 = full backstop underway). Each carries its own stated reason for holding or moving, and, when the trend is meaningful, a note on whether it's accelerating or slowing. BTC, ETH, and XRP each get a short trigger note plus a score reflecting how much that asset is actually functioning as settlement/bridge liquidity (XRP) or confirming the broader liquidity regime (BTC/ETH) right now, not just price momentum — XRP additionally gets a bridge-role read (adjacent asset through core routing asset) and whether stablecoin flow is actually routing through it or bypassing it.
The headline score is a synthesis of six domains tracked individually, each its own evidence-backed read: Treasury Market Stress, Funding Market Stress, Fed/Treasury Response, AI Economic Impulse (with a disinflationary/neutral/inflationary classification), Energy System Stress, and Global Liquidity. When populated, these appear as their own panel — useful for seeing which domain is actually driving a move in the headline number, not just that it moved.
Where present, each evaluation's supporting detail is split into three explicitly labeled tiers: observed facts (measurements, each with a source), interpretation (what those facts imply — the model's judgment), and speculation (possibilities beyond what the facts currently establish). Speculation is always labeled as the lowest-confidence tier regardless of its own stated confidence number — the label, not the number, is what governs how much weight to put on it.
Alongside the numeric band, each evaluation is classified by observed policymaker behavior — not the score alone — into one of five states: Normal, Stress Building, Controlled Stabilization, Systemic Backstop, or Monetary Regime Change. Where present, a call also states one specific, checkable claim about what a real Fed/Treasury response would look like and by when — the basis for eventually scoring whether published calls were actually followed by what they predicted.
This is an information product, not a trading system or an advisory service, and there is no human-approval step in the publishing path — a material evaluation publishes automatically the moment the three-stage process completes.
Live web research plus a FRED-sourced macro grounding context (Treasury yields, a broad-dollar proxy, high-yield credit spreads) supplied to the research stage on every cycle.