1. understand **Daniel Rodriguez by his bio in the second attached.** 2. understand his talk on health and transportation from the reflection notes from second attached. 3. understand angie's thought process from the third attached (perceive, probabilistic reason, plan) based on the talk by **Daniel Rodriguez from ops4entrep(47).pdf** 4. based on 1,2,3 write angie's 250-400 review of **Daniel Rodriguez 's talk** I apologize for the misunderstanding. After reviewing the paper more closely, I can provide a better explanation of the three-level framework in the context of rational-resource-strategy, using the SAFE (Simple Agreement for Future Equity) example as an illustration. Here's a table that encapsulates this approach: | Level | Description | SAFE Optimization Context | |-------|-------------|---------------------------| | Computational | Defines the problem and goal in terms of optimal behavior | Maximize expected utility: E[utility(i, c, M, s, f)] | | | | Where utility = founder's share × price per share | | | | Modeled using Bayesian decision theory | | Algorithmic | Specifies the process for achieving the computational-level goal | Conversational inference process: | | | | 1. Perceiving: Break down SAFE terms (investment, valuation cap) | | | | 2. Probabilistic Reasoning: Model scenarios till Series A | | | | 3. Planning: Refine terms based on ownership and dilution calculations | | Implementation | Describes how the algorithm is realized in practice | Psychological factors: | | | | - Cognitive biases (e.g., anchoring, overconfidence) | | | | - Communication strategies in negotiations | | | | Technical infrastructure: | | | | - Probabilistic programming platforms | | | | - Domain-Specific Language for financial modeling | | Section | Content | | -------------------------------- | --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- | | 1. Introduction | - Background on startup financing and SAFE agreements<br>- Importance of rational decision-making in negotiations<br>- Brief overview of the three-level framework | | 2. Theoretical Framework | - Explanation of rational-resource-strategy approach<br>- Detailed breakdown of computational, algorithmic, and implementation levels | | 3. Computational Level | - Definition of the problem in terms of optimal behavior<br>- Mathematical formulation of expected utility maximization<br>- Application of Bayesian decision theory | | 4. Algorithmic Level | - Description of the conversational inference process<br>- Breakdown of perceiving, probabilistic reasoning, and planning steps<br>- Specific application to SAFE negotiations | | 5. Implementation Level | - Analysis of psychological factors influencing decision-making<br>- Discussion of cognitive biases and negotiation strategies<br>- Overview of technical infrastructure requirements | | 6. Case Study: SAFE Negotiations | - Detailed example of applying the framework to SAFE agreements<br>- Step-by-step walkthrough of the decision-making process<br>- Comparison of outcomes with and without the framework | | 7. Discussion | - Strengths and limitations of the approach<br>- Potential extensions and future research directions | | 8. Conclusion | - Recap of key points<br>- Implications for entrepreneurs and investors | | References | List of cited works and further reading | This table illustrates how the rational-resource-strategy approach is applied across the three levels: 1. Computational Level: At this level, we define the core problem of maximizing expected utility for both founders and investors. This is expressed mathematically and incorporates Bayesian decision theory to handle uncertainty in future valuations. 2. Algorithmic Level: This level describes the process of conversational inference, which involves iteratively perceiving the situation, using probabilistic reasoning to evaluate options, and planning actions based on this analysis. In the SAFE context, this involves breaking down terms, modeling various scenarios, and refining the offer based on calculated outcomes. 3. Implementation Level: At this level, we consider both the psychological factors that influence decision-making (such as cognitive biases and negotiation strategies) and the technical infrastructure required to support the analysis (like probabilistic programming platforms and specialized modeling languages). This framework allows entrepreneurs to approach SAFE negotiations as resource-rational agents, balancing the need for optimal decisions with the constraints of limited computational resources and incomplete information. It provides a structured way to navigate the complexity of startup financing while accounting for the realities of human cognition and practical limitations. | Aspect | IV (Statistical) | IVA (Monte Carlo) | CIVA_vanilla (GPT) | CIVA (Our Approach) | | -------------------------- | ------------------------------------- | ----------------------- | ----------------------------------- | ---------------------------------------------------- | | Model | ✅ Implemented (Hierarchical Bayesian) | 🔄 In progress (Excel) | ❓ Needs configuration | 🔄 In progress (Probabilistic Program) | | Data Source | Crunchbase | Cap table | Cap table, term sheet | Cap table, term sheet | | Inference Method | MCMC (Stan) | Random sampling | LLM-based | Probabilistic programming (HFPPL) | | Handles Complex Terms | ❌ Limited | ✅ Some | ✅ Yes, but may be inconsistent | ✅ Yes, with formal representation | | Dynamicity | ❌ Static | ✅ Dynamic | ✅ Dynamic | ✅ Dynamic with explicit time-series | | Rare Event Estimation | ❌ Poor | ⚠️ Limited | ❌ Poor | ❓ Needs implementation | | Optimization Method | N/A | Grid/random search | Greedy | Bayesian optimization (ADEV_jax) | | Natural Language Interface | ❌ No | ❌ No | ✅ Yes | ✅ Yes | | Next Steps | Review code and insights | Complete implementation | Configure prompts, consider LLaMPPL | Implement HFPPL, rare event estimation, and ADEV_jax | | Aspect | IV (Statistical) | IVA (Monte Carlo) | CIVA_vanilla (GPT) | CIVA (Our Approach) | | -------------------------- | ------------------------------------- | ----------------------- | ----------------------------------- | ---------------------------------------------------- | | Model | ✅ Implemented (Hierarchical Bayesian) | 🔄 In progress (Excel) | ❓ Needs configuration | 🔄 In progress (Probabilistic Program) | | Data Source | Crunchbase | Cap table | Cap table, term sheet | Cap table, term sheet | | Inference Method | MCMC (Stan) | Random sampling | LLM-based | Probabilistic programming (HFPPL) | | Handles Complex Terms | ❌ Limited | ✅ Some | ✅ Yes, but may be inconsistent | ✅ Yes, with formal representation | | Dynamicity | ❌ Static | ✅ Dynamic | ✅ Dynamic | ✅ Dynamic with explicit time-series | | Rare Event Estimation | ❌ Poor | ⚠️ Limited | ❌ Poor | ❓ Needs implementation | | Optimization Method | N/A | Grid/random search | Greedy | Bayesian optimization (ADEV_jax) | | Natural Language Interface | ❌ No | ❌ No | ✅ Yes | ✅ Yes | | Next Steps | Review code and insights | Complete implementation | Configure prompts, consider LLaMPPL | Implement HFPPL, rare event estimation, and ADEV_jax |