# jeff ## 1. control for team! **Current H3 (β₇):** `Vagueness Γ— SeriesB Γ— High_Integration_Cost` β€” tests if reversal is stronger in hardware vs software. **Jeff's alternative (not in your model):** `Vagueness Γ— SeriesB Γ— Founder_Credibility` β€” tests if reversal only works for credible founders. These are **different mechanisms**: - **Your H3**: Reversal depends on technical uncertainty (hardware = high discovery cost) - **Jeff's idea**: Reversal depends on social capital (prior exits = trust to be vague) Your file already has `Founder_Prior_Exit` as a **control**, not a **moderator**. Jeff suggests it should be tested as a moderator instead of (or in addition to) integration cost. You could test both: - Model 2A: `Γ— High_Integration_Cost` (your current H3) - Model 2B: `Γ— Founder_Prior_Exit` (Jeff's suggestion)![[Jeff empirical guru guidance on hidden commitment cost_otter_ai (1).txt]] **Core Guidance:** - **Mechanism**: Vagueness = flexibility (no rigid expectations). Founders with **credibility signals** can be strategically vague, forcing investors to bet on the team, not the idea. - **Data**: Use existing certitude score from org science paper. Pitchbook fine. - **Model**: Simple logistic regression: - DV: `Funding_Success` (0/1) - IV: `Vagueness` (100 - certitude) - Interaction: `Vagueness Γ— SeriesB` - Controls: founder experience, team size, industry FE - **Tables**: Show progression (vagueness-only β†’ + controls) to prove robustness **Implemented:** - βœ… Logistic framework - βœ… Panel structure (A β†’ B) - βœ… Interaction term - βœ… Pitchbook data **Not Yet Addressed:** - ⚠️ Certitude measure (need to verify org science paper = LIWC) - ❌ Credibility as mechanism (Jeff emphasizes this; we test integration cost instead) - ❌ Table progression (only final models, not intermediate) - ❌ Endogeneity ("why choose vagueness?" - selection bias unaddressed) **Key Quote:** > "Strategic mechanism in being vague...force investors to invest in the team, not the idea. Lowers likelihood of getting funded, but conditional on funding, increases likelihood of delivery." **Actions:** 1. Check org science vagueness measure vs our LIWC 2. Add: `Vagueness Γ— SeriesB Γ— Founder_Track_Record` 3. Build: Model 1 (vagueness) β†’ Model 2 (+ controls) β†’ Model 3 (+ interaction) --- ## #scott (Hart Posen - Paper Structure Advisor) **Implemented:** - βœ… Formatting (no subsections in intro) - βœ… Simplified game (2-player: entrepreneur + customer) - βœ… One idea focus (strategic ambiguity) - βœ… Mean/variance separation **Not Yet Addressed:** - ⚠️ "Explain to your mom" simplicity (still technical) - ⚠️ Remove exogenous probability critique (deleted Section 1.2 but philosophy persists) - ❌ Actual intro (hypothesis.md β‰  paper intro) **Quote:** > "A paper is an argument for ONE thing, not 10." **Action:** Lead intro with "vague promises paradox", not Bayesian machinery. --- ## #unknown_speaker (from document 8 - YC/CDL pitch deck mining suggestion) **Core Guidance:** - Sample 20-30 firms from one domain (clean energy, AI drug discovery) - Mine pitch decks, code vagueness via LLM - Track Series A β†’ B with CrunchBase - Model: Strip to 1/3 symbols, derive optimum, explain deviations (agency/bounded rationality) **Status:** Partially implemented (Pitchbook replaces YC/CDL, but same logic) --- ## Combined Gaps **Critical:** 1. Jeff + Unknown both want **credibility** tested, not just integration cost 2. Model too complex (9 variables) - contradicts "proof of concept" guidance 3. No table progression (vagueness β†’ + controls β†’ + interaction) **MVP Priority:** 1. Verify vagueness measure consistency 2. Test credibility alongside integration cost 3. Simplify variable count (drop less critical controls)