Flagship prototype
Great Bear Football
First public proving ground
Football makes the interaction inspectable: a proposed tactical action, adversarial adaptation, failure mode, robust action within the authored case, trigger, and downside.
Counterfactual intelligence
Test the strategy before reality does.
Great Bear is building systems that turn a proposed action into plausible and adversarial futures, expose where a plan fails, and return a contingent move with a trigger and downside.
Strategy stress testing and scenario analysis. Football proves the interaction in public. Finance is the second quantitative application.
Football proof
A tactical case that can be inspected
This is an authored interaction proof, not a model result. The shift stays inside the declared assumptions and answers the weak-side failure mode.
Illustrative tactical case · Authored assumptions · Tier D — unresolved
CURRENT PLAN — 4-3-3 high press
OPPONENT ADAPTATION — No. 6 drops into the first line
FAILURE MODE — weak-side access opens after the first jump
ROBUST ACTION — shift to 4-4-2 pressure
TRIGGER — the No. 6 receives between the centre-backs
DOWNSIDE — more space behind the wide midfielder
This is not a model result. The adjustment is recommended within this authored case and remains unresolved until valid evaluation exists.
Open the Football pageStrategy workflow
One prediction is not a strategy.
A strategy system must test what happens after you intervene, how another agent may adapt, and which response still works when the original assumptions fail.
Define → Branch → Shear → Compare → Act → Resolve
Define
Branch
Shear
Compare
Act
Resolve
Shear Test
A Shear Test is an adversarial perturbation of a declared assumption.
Shear Profile
Prefer a Shear Profile over any universal single score until a normalization and calibration protocol exists.
Robust means scoped
Robust means robust within the declared branch family, assumptions, constraints, and evidence tier.
Applications
Football first, finance second, one core system.
Research prototype · Design-partner program
Great Bear Finance
Second quantitative application
The proposed Portfolio Stress Lab would compare declared hold, rebalance, cash, or constrained-hedge actions across disclosed synthetic Scenario Branches, then report modeled tail loss, constraint failures, implementation cost, and ranking stability. Research only: No execution · No live feeds · No return or performance claims · Not investment advice.
Shared system under development
Great Bear Strategy Engine
One reusable case system
The Strategy Engine is the shared state–action–prediction–outcome contract behind the public proof, future quantitative prototypes, and resolution tooling.
Decision moat
The moat is resolved decision evidence, not scenario volume.
Decision Fidelity is a declared evaluation protocol and metric vector, not a single confidence score. It records what the case asked, how actions were ranked, and how the resolution was allowed to compare them.
Observing the chosen action usually does not reveal what unchosen actions would have done.
ranking agreement
decision regret
coverage
calibration
stability
The Decision Graph is the intended compounding asset: permissioned, resolved state–action–prediction–outcome evidence with provenance and evidence tier. The authored Football case is Tier D, and the resolution tooling is in development.
Read the evidence methodologyAgent interfaces
One case contract can serve people first, agents later.
A versioned Strategy Case contract is intended to serve web UI, manual fields, file inputs, future agent tools, and ChatGPT/Claude app surfaces without changing the underlying decision object.
web UI
manual fields
file inputs
future agent tools
ChatGPT/Claude app surfaces
These integrations are a product and interface direction — not callable interfaces today, not MCP surfaces, not delegated execution, and not host-platform approval. Human approval remains the boundary.
Explore the Strategy Case contractPilot Wave
Pilot Wave is a benchmark program, not the moat.
Pilot Wave is a provider-neutral benchmark program. Strong classical baselines come first; where appropriate, selected discrete workloads are also evaluated with annealing, hybrid, and gate-model methods. Provider access implies no partnership or endorsement.
Methods compared
classical baselinesannealinghybridgate-model
Measures reported together
feasibilitysolution qualitysolution diversitystabilitytotal wall timeoverheadscostdownstream decision impact
There is no quantum-advantage claim. PyTorch MPS on Apple Silicon is a local classical compute path, not a quantum provider.
Read the Pilot Wave methodologyInvestor path
The investor brief starts with evidence.
Read the full category, wedge, moat, and research boundary — then inspect the public Football case that anchors the current story.