The LLM's natural behavior: Generate risk assessments, due diligence questions, skeptical
analysis, and "things to investigate" - because that's what a thorough analyst would do.
The business reality: Investment memos serve multiple audiences over time:
- Internal discussion (risks useful here)
- Decision preservation (judgment, not uncertainty)
- Syndicate building (promoting the deal)
- LP review (demonstrating conviction, not doubt)
A memo full of "we're not sure about X" and "key risk: Y" and "due diligence priority:
investigate Z" undermines all but the first use case.
Proposed solution structure:
- Public memo → Clean, conviction-driven, risks acknowledged but contained
- Private diligence folder → All the skepticism, risk assessments, investigation checklists,
hanging questions
So the flow becomes:
0-deck-sections/ → Extract from deck
1-research/ → Web research
2-diligence/ → Express ALL skepticism, risks, questions (cathartic)
3-sections/ → NOW write conviction-focused memo (skepticism already externalized)
4-final-draft.md → Assembled from 3-sections/
The LLM's natural behavior: Generate risk assessments, due diligence questions, skeptical
analysis, and "things to investigate" - because that's what a thorough analyst would do.
The business reality: Investment memos serve multiple audiences over time:
A memo full of "we're not sure about X" and "key risk: Y" and "due diligence priority:
investigate Z" undermines all but the first use case.
Proposed solution structure:
hanging questions
So the flow becomes:
0-deck-sections/ → Extract from deck
1-research/ → Web research
2-diligence/ → Express ALL skepticism, risks, questions (cathartic)
3-sections/ → NOW write conviction-focused memo (skepticism already externalized)
4-final-draft.md → Assembled from 3-sections/