How Anthropic ships the #1 LLM in the World — with Katelyn Lesse and Angela Jiang
About this episode
From the show’s notesAn agent’s sandbox can die without taking the whole agent with it. That’s the idea behind “the brain and the hands,” Anthropic’s architecture for Claude Managed Agents.
Now that the dust has settled on Opus 5.5, it is universally agreed to be the best model in the world. In this conversation with Katelyn Lesse and Angela Jiang, Anthropic’s Heads of Engineering and of Product, goes inside the systems that turn model capabilities into products developers can use.
Read the show’s notes in full
We cover:
- Why Anthropic separated the agent harness, session state and execution environment. - Letting agents use credentials without putting those credentials inside their sandboxes. - Rebuilding the Messages API from Python to Rust with Claude—and verifying the rewrite with shadow traffic. - Preventing model distillation while preserving the developer experience. - Memory, “dreaming,” agent evaluations and the work still left for human engineers. - How Anthropic’s own teams use Claude while keeping responsibility for their code and writing.
The episode opens with swyx’s look at Anthropic’s S-1: revenue per employee, compute efficiency, and the improvement and investment loops behind the business.
**Related links**
- [Claude Opus 5.5](anthropic.com/claude-opus-5-5) — Performance, pricing and safeguards.
- [The brain and the hands](
anthropic.com/research/managed-…) — The Managed Agents architecture.
- [How Anthropic contains Claude](
anthropic.com/research/how-we-c…) — Isolation, credentials and network controls.
- [Harnesses for long-running agents](anthropic.com/research/effectiv…) — Maintaining progress across context windows.
