Example 01
Internal research system
Turn available data into researchable equity theses
Agent teams search the data landscape, improve promising ideas through repeated review, and return a short list worth deeper research.
The problem
A familiar bottleneck
What the system does
A simpler way through it
- BriefCurrent stage
- GenerateUpcoming
- ValidateUpcoming
- ImproveUpcoming
- IntegrityUpcoming
- OutputUpcoming
Power demand, grid investment, and project development
- Research field
- U.S. power and infrastructure equities
- Research question
- Where could changes in power demand, grid investment, or project development create an overlooked equity-research opportunity?
- Company universe
- U.S.-listed utilities, power producers, infrastructure developers, and electrical-equipment companies
- Research horizon
- 2–5 months
- Review cadence
- Weekly
Permitted sources
- Approved internal datasets
- Public filings and earnings materials
- Government and regulatory datasets
- Public ISO and RTO records
- Public company and project disclosures
Constraints
- Public or approved data only
- Use information available at the time
- No MNPI
- No fabricated or guessed values
- Must be testable across multiple companies
- Data and engineering cost must be reasonable
- Every idea must include a failure condition
Useful output
What the research team gets
- A broader scan of useful internal and public data
- 50 candidates considered across five improvement cycles
- A transparent record of why ideas advanced or were archived
- Three evidence-backed theses with data and testing plans
- Costs, limitations, competing explanations, and failure conditions
- A ranked research queue ready for human review
Turn available data into researchable equity theses. Step 1 of 6: Set the research brief. The orchestrator defines what the teams may research, which sources they may use, and what a useful idea must contain.