Anthropic agent engineer:

“Don’t give the agent everything. Give it the right context, at the right time.”

In this 12-minute video, he explains what actually makes Claude agents work in production:

>> why an 80%+ cache hit rate matters

>> how too much context quietly ruins an agent

>> loading tools only when they are needed

>> using an Opus advisor for harder decisions

>> compaction for agents that run much longer

This is the engineering behind reliable agents.

Not a bigger prompt. A cleaner system.

watch the session, bookmark it.

then read my guide below to build your first Claude agent from scratch.