The calibrated judgment layer for consequential action.
Warm Winter is not a tool, an SDK, or an app — those are how it enters. It's the verified record of what to do, and what happens when you do it, that machines and people consult before they act. We answer one question, everywhere it's asked: is this trustworthy enough to act on — and what follows if you do?
One engine, two faces.
To an AI agent, Warm Winter is the gate that decides whether an action is trustworthy enough to run unattended. To an operator, it's the engine that shows the likely outcome of a decision before they commit. It's the same question both times — is this trustworthy enough to act on, and what happens if you do? — answered the only honest way: scored against what actually happened.
Both sides exist for the same reason: confidence and correctness come apart. An agent — or a person — can sound exactly as sure when it's wrong as when it's right. Warm Winter's only job is to tell the difference, and to keep proving it against reality. That's the whole company — not an AI tool or a business tool, but the calibration layer beneath consequential decisions, human or machine.
Infrastructure with a calibrated opinion — sitting beneath decisions, not as a product on top of them.
The category isn't “another SaaS tool.” It's FICO (the score every lender runs), Bloomberg (the terminal every desk consults), the clearing layer, the audit layer — small companies that enormous flows route through. Plumbing that holds an opinion about what to trust is rarer, and more defensible, than plumbing.
Three expressions of the same thing.
The whole company is a single calibrated judgment engine — domain-blind by design — plus a single compounding asset: a proprietary, ever-growing record of (situation → decision → outcome) tied to reality. That one engine expresses itself in three forms, co-equal in the vision and strictly sequenced in execution — one wedge at a time, depth before breadth.
The AI trust layer
“Is this agent or model action trustworthy enough to act on — or should it escalate?”
The calibrated gate over agentic and compute decisions. Our wedge, and where the company is today.
The operator layer
“What do I do, and what happens if I do?”
Judgment for the humans who run businesses — pricing, operations, risk, allocation. The engine's proof on human-stakes decisions, and the bridge to judgment in everyday life.
The physical layer
“Own the hand.”
Judgment that commands physical production and assets directly, gated by the competence frontier as a safety interlock. The deepest expression — where it earns the right not just to advise, but to act.
The discipline that keeps three expressions from becoming three half-built products: the vision is wide; the focus is narrow. We run one wedge at a time and earn each expansion with verified depth in the prior. The grand company and the narrow wedge are the same plan viewed from two distances.
Three horizons, one throughline.
The arc widens with time; it does not erode. Each horizon is earned, not announced — this is the destination, drawn honestly, not a claim about today.
Own the gate over agentic and compute decisions. Wedge = compute; entry = developers running agents. Here today — live in prod.
Energy → supply chain → finance, and operators at scale. The same engine becomes the allocation-and-trust brain across coupled resources — compute ↔ energy ↔ capital.
The verified record becomes the check every operator, institution, and agent runs before acting — consulted everywhere, by humans and machines. Ambient is the destination, never the doorway.
Own the judgment, not the interface
The face of AI will be built and commoditized by the largest companies on earth. We don't race them for it. We're the judgment inside — the check the agentic world calibrates against. Own the hand; cede the face.
Earn ubiquity by verified depth
Judgment earns the right to be everywhere only by being verifiably right somewhere first. Depth of verified outcomes beats breadth, always.
Grow with AGI, not against it
The cost of a wrong decision rises as intelligence gets stronger — so the smarter the systems, the more valuable judgment about judgment. As agency spreads to billions of agents, every one needs a trust gate.
The asset can't be downloaded. It only deepens.
Models get smarter and sensors get cheaper — those commoditize. The compounding asset is a five-decade record of what actually happened when people and machines acted, made legible by the competence frontier: per-cell verified, provisional, or ungrounded — never a blended global accuracy number. The most valuable thing it builds is the part that says “I don't know this yet.”
At its deepest, Warm Winter is a cognitive prosthesis for judgment — externalizing calibration, enforcing the discipline of scoring yourself against reality. And the competence frontier is a candidate answer to the question of the century: how does humanity stay in calibrated control of systems that become smarter than us? By granting autonomy only where trust has been verified — and withholding it everywhere else.
The interface of AI will be everywhere and free. The judgment underneath it — grounded in reality, honest about its own edges — won't be.
AI is about to act on our behalf everywhere — in our code, our money, our infrastructure, our lives. What unsettles me isn't that it'll be wrong sometimes; it's that it sounds exactly as sure of itself when it's wrong as when it's right. Warm Winter is my answer: a layer that earns trust the only honest way — by remembering what actually happened every time it made a call, and being willing to say it doesn't know. Built solo, in the open, one verified outcome at a time.
— Enrique Vigil, founder · LinkedIn · a direct line
Start with the part that's live.
The vision is the cathedral; the trust gate for AI builders is the first stone — and it's in your hands today.