496. How Model Progress Shifts the Goalposts, Why The Death of Software Is Overstated, and How to Diligence Hypergrowth Without Getting Burned (Jacob Effron)

What It Covers

Jacob Effron from Redpoint discusses AI model progress, evaluating hypergrowth startups in the AI era, vertical AI applications in healthcare and legal, agentic AI development, and why traditional T2D3 growth metrics no longer indicate exceptional product-market fit.

Key Questions Answered

• •AI Growth Benchmarks: Traditional triple-triple-double-double-double growth is no longer exceptional for AI companies. Reaching 100 million ARR in under three years has become the new standard for top-tier AI product-market fit, given how rapidly AI solutions are being adopted across industries and enterprises today.
• •Model Progress Focus: Foundation models show strongest improvements in coding and math through reinforcement learning, but future progress will be domain-specific rather than universal. Application companies now develop their own evaluation benchmarks that matter more than standardized tests for measuring real-world model performance in specialized verticals.
• •Vertical AI Defensibility: Data moats are overestimated in AI. Fine-tuning and reinforcement learning require less proprietary data than expected. Defensibility comes from connecting ecosystem stakeholders, strategic partnerships, funding velocity enabling broader product development, and deeply understanding domain-specific evaluation metrics that drive product quality in healthcare, legal, and logistics.
• •Diligencing Hypergrowth: When evaluating AI companies growing at extreme rates, prioritize team velocity to adapt quickly, identify persistent customer needs that survive 10x model improvements, and focus on landing with customers today to evolve together rather than predicting future needs from an ivory tower without deployment learnings.