Podcast insights
Building AI agents with Claude — Anthropic's engineering notes
Notes from Anthropic's official Claude talks on building real agents — straight from their engineers. Quotes are verbatim; sign in to read the full transcripts.
From 6 episodes on PodLearn
Pick the right primitive: tool, skill, or subagent
We want to make sure that we're using the right agentic primitives at the right time. So when do you use a tool, when do you use a skill, and when do you use a subagent?
Tool, skill, or subagent? Decomposing an agent that outgrew its prompt · Claude (Anthropic)
Agents degrade as they outgrow their prompt
Your system prompt had grown to become several hundred lines long. You have dozens of tools and subagents… and because of the complexity, you've started to see regressions in the areas that your agent was previously accelerating in.
Tool, skill, or subagent? Decomposing an agent that outgrew its prompt · Claude (Anthropic)
Don't just bolt on capabilities — modernize the architecture
We've essentially bolted capabilities onto our agent over time without modernizing our architecture. This complexity has started to cause some problems.
Tool, skill, or subagent? Decomposing an agent that outgrew its prompt · Claude (Anthropic)
Episodes this draws from
- Tool, skill, or subagent? Decomposing an agent that outgrew its prompt
Claude (Anthropic)
- Ship your first Managed Agent
Claude (Anthropic)
- Designing with Claude: From prompt to production
Claude (Anthropic)
- Agents that remember
Claude (Anthropic)
- Building the best agentic analytics harness
Claude (Anthropic)
- Build a proactive agent workflow with Claude Code
Claude (Anthropic)
Full transcripts are free to read with a PodLearn account.
FAQ
When should you use a tool vs. a skill vs. a subagent?
Anthropic's engineers frame it as choosing the right agentic primitive for the job — tools for actions, skills for reusable on-demand instructions, and subagents for isolated context — rather than piling everything into one orchestrator.
Why do AI agents get worse over time?
Because capabilities get bolted on without modernizing the architecture: the system prompt balloons, tools and subagents multiply, and the agent starts to regress where it used to excel.
Where can I read these Anthropic agent talks?
On PodLearn — full verbatim transcripts of Anthropic's Claude talks, searchable and free to read with an account.
Read every word — free
Every quote above links to its full, searchable transcript on PodLearn. Create a free account and read, search, and turn any episode into a lesson.