CLAUDE CODE AGENTS: ADVANCED ORCHESTRATION - Chapter 1, Exercise 3
Why Fine-Tuning Falls Outside This Course's Scope
====================================================================================

QUESTION:
A learner asks whether this course will teach them to fine-tune a model so
it behaves better as part of an agent pipeline. Using this chapter's own
warning box, explain why that falls outside this course's scope, and what
it teaches instead for improving agent behavior.

SOLUTION / EXPLANATION:

This chapter's warning box names fine-tuning directly as one of three
things deliberately out of scope: fine-tuning changes a model's own
underlying weights through additional training, which is a fundamentally
different kind of technique from anything else covered in this course.
Every technique in this course - the Agent SDK, orchestration patterns,
specialized agent design, handoffs, debugging, cost tradeoffs - works with
models exactly as they already are, shaping behavior through prompting,
tool design, and coordination between agents, not through retraining the
model itself.

The reason this is excluded isn't that fine-tuning is unrelated to agent
behavior in principle - it's that it belongs to a genuinely different
discipline (machine learning model training) requiring its own separate
treatment, rather than a rushed addition to a course about designing and
orchestrating agents using models as a fixed, unchanging component.

What this course teaches instead, for improving how an agent behaves within
a pipeline, are the actual techniques it covers directly: writing a
well-designed system prompt (the same discipline from Fundamentals'
Chapter 2, applied at pipeline scale), choosing the right tool allowlist for
each agent's actual job, structuring how agents hand off context to each
other cleanly (Chapter 7's own topic), and debugging and evaluating where
in a chain behavior actually goes wrong (Chapter 8's own topic) so it can
be corrected through better prompting or orchestration design - not through
retraining the underlying model.

--------------------------------------------------------------------------
WHY THIS WORKS AS AN ANSWER:
It explains specifically why fine-tuning is a different kind of technique
(retraining model weights vs. everything else in this course, which works
with a fixed model), and names the concrete alternative techniques the
course actually teaches for improving agent behavior without retraining.