Generated Skill

Problem: [Startup pitch/technology description]

Rapid VC Hash (5-min evaluation):
1. Physics check: Is this computationally/physically feasible?
2. Problem value: High value + high frequency + increasingly painful?
3. Causal chain: Technology → Cost → Behavior → Organization → Market
4. Kill test: Where does this most likely die?
5. Anti-consensus angle: What counternarrative exists?

Decision: [Signal/Noise] + [Next phase warranted: Y/N]

Phase 0: Background Daemon (Always Running)

• Track compute cost curves, research breakthroughs, infrastructure pricing
• Update technology timeline hypotheses and industry migration maps
• Monitor open-source ecosystems and boring-but-growing tools

Phase 1: Opportunity Capture

Progress:
- [ ] Source identified (cold email, technical community, academic paper, etc.)
- [ ] Quick filter: Anomaly signal or noise?
- [ ] Decision: Deserves cognitive cycles Y/N

Phase 2: Rapid Compression (30-60 min)

Progress:
- [ ] Map to 2-3 known paradigms
- [ ] Identify single critical variable being bet on
- [ ] Determine most likely failure mode
- [ ] Generate cognitive hash summary

Phase 3: Deep Modeling (1-3 weeks)

Progress:
- [ ] Decompose problem space completely
- [ ] Draw full causal chains (1st, 2nd, 3rd order effects)
- [ ] Interview 5-10 domain experts (unrelated to company)
- [ ] Actively attempt to kill the idea
- [ ] Document why idea survived kill attempts

Phase 4: Investment Decision

Progress:
- [ ] Record investment thesis
- [ ] Document causal chain being bet on
- [ ] Identify fatal assumption (if wrong, money should die)
- [ ] Assess mistake survivability
- [ ] Make binary decision: Invest/Pass

Example 1: AI Startup Evaluation Input: "We're building AI agents for customer service" Output:

• Physics check: ✓ (LLM inference costs dropping exponentially)
• Problem map: High frequency, increasingly painful as scale grows
• Causal chain: Model capability ↑ → Inference cost ↓ → Human labor arbitrage → Organizational restructuring
• Kill test: Dies if: (1) Hallucination problem unsolved, (2) Regulation kills automation, (3) Incumbents integrate faster
• Anti-consensus: Everyone's building this - what's differentiated?
• Hash: Noise unless solving specific technical bottleneck or unique distribution

Example 2: Infrastructure Technology Input: "New database architecture for real-time analytics" Output:

• Physics check: What's the fundamental computational limit being approached?
• Problem decomposition: Engineering problem (money+time solves) vs information-theoretic limit
• Causal model: Performance improvement → Cost reduction → New application classes possible
• Expert validation: Talk to 5 database engineers, 3 data scientists, 2 infrastructure VCs
• Kill attempts: Existing solutions + cloud scaling, vendor lock-in risks, market timing

Technical Intuition (Without Coding)

• Focus on order-of-magnitude understanding: latency, bandwidth, compute, energy costs
• Distinguish real bottlenecks from problems Moore's Law will solve
• Judge feasibility through physics/information theory constraints

Problem-First Thinking

• Prioritize: What problem + Why now + Who cares desperately
• Ignore: Solution elegance, architecture beauty, technical sophistication
• Remember: VCs buy irreversibly expanding problems, not solutions

Causal Chain Modeling

• Always map: Technology → Cost → Behavior → Organization → Market structure
• Track empowerment shifts: who gains/loses power
• Model 2nd and 3rd order effects, not just direct impacts

Anti-Consensus Calibration

• Construct counterexamples to every hot narrative
• Ask "why now?" for ignored directions
• Early + correct + temporarily lonely > popular consensus

Avoid Solution-Space Obsession

• Don't get seduced by technical elegance
• Don't optimize for architectural purity
• Don't mistake implementation complexity for value

Don't Ignore Human Factors

• Technology gets copied, humans don't
• Cognitive collapse kills more startups than technical failure
• Look for non-instrumental problem obsession in founders

Resist Narrative Self-Intoxication

• Record specific falsifiable assumptions
• Build kill-switches into investment thesis
• Distinguish between wanting something to be true and evidence it's true

Post-Investment Restraint

• Don't become product advisor or management coach
• Intervene only at: cognitive blind spots, critical decisions, narrative distortions
• Function as external cognitive calibration, not hands-on guidance