# How We Build Effective Agents: Barry Zhang, Anthropic

> For agents: start with the [agent guide](https://ratstack.sh/llms.txt). Every page is Markdown by default; add `Accept: text/html` for HTML.

[AI Engineer](https://www.youtube.com/@aiDotEngineer) · Published 2025-04-04 · 15:09 · A tier · Agents and MCP

[Watch on YouTube](https://www.youtube.com/watch?v=D7_ipDqhtwk) · [All RatTube videos](/tube)

## Summary

Barry Zhang argues that an agent is useful when a task needs exploration rather than a fully known decision tree. He asks builders to consider both task complexity and the value of spending model calls on that exploration.

He models agents as models using tools in a loop, operating within an environment. His design advice is to keep that loop simple and inspect the context the model actually sees. Reviewing tool descriptions and execution histories can reveal why an agent made a poor choice.

## Section notes

- [0:32](https://www.youtube.com/watch?v=D7_ipDqhtwk\&t=32s): The talk introduces three themes: selective use, simplicity, and the agent's perspective.
- [2:54](https://www.youtube.com/watch?v=D7_ipDqhtwk\&t=174s): Ambiguity and task value help decide whether an agent is appropriate.
- [5:35](https://www.youtube.com/watch?v=D7_ipDqhtwk\&t=335s): An agent is described as a model using tools in a loop.
- [8:13](https://www.youtube.com/watch?v=D7_ipDqhtwk\&t=493s): Inspecting the context window helps explain an agent's decisions.
- [10:51](https://www.youtube.com/watch?v=D7_ipDqhtwk\&t=651s): Tool descriptions and execution histories offer material for improving decisions.

## Key concepts

No matching wiki page yet.

## Our take

Barry Zhang on three ideas from Anthropic's Building Effective Agents: don't build agents for everything, keep them simple, and think like your agent. Fifteen minutes that will change how you design tools.

## Related videos

- [Code Mode: Let the Code do the Talking - Sunil Pai, Cloudflare](/tube/code-mode-let-code-do-talking)
- [All Code Is Slop (Except Mine) — Lauren Tan, SpaceXSI](/tube/all-code-is-slop-except-mine-lauren-tan)
- [The Future of MCP — David Soria Parra, Anthropic](/tube/future-of-mcp-david-soria-parra)
- [Don't Build Agents, Build Skills Instead – Barry Zhang & Mahesh Murag, Anthropic](/tube/dont-build-agents-build-skills)

Summary based on the transcript. Written 2026-11-15.
