Skip to content
Lenny's 🏦 Vault

Three-Part Equation for AI Product Utility

by Mike Krieger on June 5, 2025

Mike Krieger's "Three-Part Equation for AI Product Utility" reveals how Anthropic approaches building valuable AI products by focusing on three critical components that must converge for maximum effectiveness.

The Three-Part Equation for AI Product Utility

  • Model intelligence + Context/Memory + Applications/UI = Useful AI Product
    • "My fake equation for utility of AI products is three part: one is model intelligence, the second part is context and memory, and the third part is applications and UI. You need all three of those to converge to actually be a useful product in AI."

1. Model Intelligence

  • The core capabilities of the underlying AI model
  • Driven by research teams focused on improving model performance
  • Measured through benchmarks (e.g., "We're at about 72% now with the new models and we were at 50% when [Dario] made that prediction")
  • Continues to scale and improve predictably over time

2. Context and Memory

  • The ability to access and utilize relevant information
  • Critical differentiator between good and bad responses
    • "The difference between the right context and not is entirely the difference between a good answer and a bad answer"
  • Solved through protocols like MCP (Model Context Protocol)
    • Allows models to access external data sources, tools, and systems
    • Creates standardized ways for AI to interact with various applications
    • Enables composability across different systems

3. Applications and UI

  • How users discover and interact with AI capabilities
  • Focuses on making integrations discoverable and usable
  • Creates repeatable workflows around AI capabilities
  • Determines whether users can effectively leverage the model's intelligence

Implications for Product Development

  • Bottlenecks shift from code generation to higher-level concerns:

    • "We really rapidly became bottlenecked on other things like our merge queue"
    • "Over half of our pull requests are Cloud Code generated probably at this point. It's probably over 70%."

  • Product teams should focus on:

    • Strategy: "Figuring out what to build and how to win in the market"
    • Comprehensibility: "Making it easier to help people understand how to leverage the power of these tools"
    • Possibility expansion: "Opening people's eyes to the potential of these sorts of things"

  • The future of AI product development involves:

    • Everything becoming accessible through protocols like MCP
    • Models having agency to interact with various systems
    • "Everything is scriptable and everything is composable and everything is usable identically by these models"

  • Measuring success requires new metrics beyond engagement:

    • "Did it actually help you get your work done?"
    • "Claude helped me put together a prototype the other day that saved me literally probably six hours and it did in about 20-25 minutes"
    • Focus on "unlocking creativity" and giving people "more space in their lives"
View full episode