Videos · Architect Foundations · task 1.1
Agentic loops
Five short videos on how an agentic loop works. Who does what, how it knows when to stop, how tool results get back to Claude, workflow or agent, and the exam trap of stopping too early.
5 parts · 11 min
- 1 What is an agentic loop? 2:12
Chapters
- 0:00 Welcome
- 0:10 The question
- 0:27 Two players and a loop
- 0:56 Round one
- 1:19 Round two
- 1:35 Round three, and out
- 1:51 Recap
- 2 How does an agentic loop know when to stop? 2:08
Chapters
- 0:00 Welcome back
- 0:15 One field decides
- 0:29 The two you'll see most
- 0:51 Cut off
- 1:03 Tools that run on Anthropic's side
- 1:24 For the exam, and for your code
- 1:46 Recap
- 3 How tool results get back to Claude 1:51
Chapters
- 0:00 Welcome back
- 0:15 Why the whole conversation
- 0:32 Two messages per round
- 0:59 Several tools in one reply
- 1:13 What goes wrong if you drop rounds
- 1:27 Recap
- 4 Workflow or agent: who picks the next step? 1:36
Chapters
- 0:00 Welcome back
- 0:15 Two ways to build it
- 0:30 A result nobody planned for
- 0:59 So only Claude can say it's done
- 1:15 Recap
- 5 The agentic loop that stops too early (and the exam trap) 2:44
Chapters
- 0:00 Welcome back
- 0:15 The answer that stops halfway
- 0:29 Wrong signals one and two: the words
- 0:57 Wrong signal three: the count
- 1:27 The fix
- 1:54 The exam
- 2:21 Recap of the series