Invest-GT is easiest to operationalize as a staged research pipeline rather than as a monolithic model. The goal is to move from explicit assumptions to reproducible runs and structured evaluation.
Define the baseline world representation used by all downstream work:
- actors, institutions, alliances, blocs
- macro variables and policy variables
- event types and constraints
- asset universe or allocation targets
- scenario assumptions and time horizon
Primary workspace:
pipeline/00_inputs/
Write down the claims the system is supposed to test before building runtime logic.
Primary workspace:
pipeline/01_hypotheses/hypotheses/
Specify actors, incentives, constraints, beliefs, and strategy spaces.
Primary workspace:
pipeline/02_agents/agents/
Define payoff structures, update rules, interaction forms, and propagation logic.
Primary workspace:
pipeline/03_models/models/
Connect the separate components through explicit interfaces and configuration rules.
Primary workspace:
pipeline/04_runtime/runtime/configs/
Execute scenarios, sweeps, and experiment variants.
Primary workspace:
pipeline/05_simulations/simulations/
Compare outputs against consistency checks, benchmarks, and historical anchors.
Primary workspace:
pipeline/06_tests/tests/notebooks/
Translate outputs into usable research findings and limitations.
Primary workspace:
pipeline/07_reports/reports/docs/
- define or update the world state in
pipeline/00_inputs/ - define a scenario or research question
- write a hypothesis in
pipeline/01_hypotheses/ - specify actors and decision logic
- formalize the interaction model
- assemble and run the runtime
- test and interpret the results
- summarize the run and its limitations
This project sits at the intersection of politics, strategic interaction, and investment behavior. That work becomes hard to reason about when assumptions, scenarios, model logic, and outcomes are mixed together. A pipeline-first structure keeps those concerns legible and versionable.