Bret Taylor of Sierra on AI agents, outcome-based pricing, and the OpenAI board

By Stripe

Categories: Product, Startup

Summary

OpenAI's new AI agents are surprisingly crude yet effective, using 'scribbles in the margin' for memory - a more efficient 'harness engineering' approach than polished mainstream apps. By mimicking engineers' own code-centric workflows, these agents are pioneering a new path for AI across industries.

Key Takeaways

  1. AI agents don't need polished UIs to be effective - open-source projects like OpenClaw, with 'janky' memory features, can outperform slick mainstream apps.
  2. Harness engineering - building a robust 'harness' around AI agents to handle tasks like documentation and testing - may be a more efficient path than building 'fancier' agents from scratch.
  3. Software engineers' own code-centric workflows and tools (like grep) provide a surprisingly effective 'harness' for general-purpose AI agents to work within.
  4. AI agents may be more useful with a 'directory of just everything you've ever done' than fancy vector databases, mirroring how human memory actually works.
  5. The rapid evolution of AI agents over just 4 months shows how quickly this technology is progressing - far outpacing typical tech timelines.
  6. AI agents may be best suited for tasks with clear context and feedback loops (like coding), rather than broad information tasks (like writing a company report).

Topics

Transcript Excerpt

Bret Taylor is the ultimate Silicon Valley veteran. He was one of the creators of Google Maps, invented the Like button, was co-CEO of Salesforce. He pushed through Elon's acquisition of Twitter when he was on the Twitter board. He's now the chairman of the OpenAI board. And his day job is founder and CEO of Sierra, which is bringing AI to customer service. He's one of the smartest people I know on the topic of how AI is changing established companies. Cheers. Cheers. Most important question, ha...