Engineering first
Infrastructure as code, automated checks, and review for every change — including configuration that defines risk limits.
About
A StreamPod research project.
MAA exists to explore how AI agents can support market research without sacrificing rigor, transparency, or control.
Mission
AI agents are good at gathering, summarizing, and cross-referencing information. They are also nondeterministic, can be wrong with confidence, and are vulnerable to manipulation through the content they read. Markets reward neither carelessness nor overconfidence.
MAA is a research project within StreamPod. Its data layer is the Unified Broker Interface (UBI): one interface across brokers, reached through each broker’s MCP server, so agents and researchers get consistent, least-privilege access to market data.
MAA is an attempt to take the useful parts of agentic AI and surround them with engineering discipline: reproducible experiments, point-in-time data, a deterministic policy layer, and honest reporting about what has and has not been shown.
The goal at this stage is understanding, not returns. The project publishes methodology and research notes so its approach can be examined and corrected.
Technical approach
Determinism where it matters
Models are used for research assistance; anything that gates an action is deterministic and testable.
Skepticism as a feature
Negative results and failed experiments are recorded alongside successes. A result that cannot be reproduced is treated as a lead, not a finding.
Open about limits
Claims are limited to what has been demonstrated. Where something is a design goal rather than a built capability, the site says so.
What MAA is not
- Not an investment fund, advisor, broker, or trading signal service.
- Not accepting or soliciting investment capital.
- Not a source of investment advice or recommendations.
- Not trading live, and not publishing performance results.