Orientation Exercises¶

These exercises make the study method observable before Module 01 asks you to change code. Work against the tracked Module 01 state, not the live endpoint, unless an exercise explicitly asks you to compare them.

Exercise 1: Choose an honest proof boundary¶

Starting context¶

A teammate makes each of these claims:

“Module 01 rejects True as a chunk size.”
“Every tracked module state matches its generated history worktree.”
“The live capstone remains healthy after a cross-package refactor.”

Objective¶

Choose the smallest suitable command for each claim from:

capstone-foundation-proof
history-verify
capstone-test
test

Constraints¶

Do not choose a broader command merely because it contains the narrower test.
Distinguish behavioral evidence from generated-state integrity.
State what each selected command cannot prove.

Expected evidence¶

Produce a three-row table with the claim, command, why it is sufficient, and proof limit.

Acceptance checks¶

The exact chunk-size claim uses the tracked Module 01 learning proof.
The state-coherence claim uses history verification rather than application tests.
The cross-package live refactor uses at least the live capstone suite.

Earlier contract to preserve¶

The capstone remains cumulative: a narrow Module 01 proof must not silently import the live Module 10 implementation.

Exercise 2: Route a focused failure¶

Starting context¶

Imagine the foundation proof fails here:

assert "".join(chunk.text for chunk in chunks) == cleaned.abstract

You have not changed embedding, deduplication, file loading, or the shell.

Objective¶

Write a failure-investigation route that begins with the smallest owning surface.

Constraints¶

Name the first source function and file to inspect.
Explain why at least two other files are not the first owner.
Give one plausible defect and one counterexample that would expose it.
Do not propose a fix until you can state the broken law.

Expected evidence¶

Provide:

the law in plain language
the first source owner
one concrete input, expected chunks, and reconstructed text
the focused command you would rerun

Acceptance checks¶

The route starts at chunk_doc in the Module 01 pipeline_stages.py.
The counterexample checks both text coverage and offsets.
The rerun remains bounded to capstone-foundation-proof.

Earlier contract to preserve¶

Cleaning remains deterministic; a chunking investigation must not alter the cleaned document to make the assertion pass.

Exercise 3: Separate the first state from the endpoint¶

Starting context¶

The live capstone contains packages for results, policies, effects, async coordination, interop, and review evidence. None of those packages appears in the tracked Module 01 source tree.

Objective¶

Explain why teaching Module 01 from the live package would distort the application delta.

Constraints¶

Inspect both capstone/module-reference-states/module-01/src/funcpipe_rag/ and capstone/src/funcpipe_rag/.
Identify two live concerns that belong to later modules.
Identify one foundation law that the live endpoint must still preserve.
Do not argue that later code is bad merely because it is more complex.

Expected evidence¶

Write a short comparison with these headings:

pressure visible in Module 01
later pressure visible only in the endpoint
preserved foundation law
teaching cost of starting from the endpoint

Acceptance checks¶

The comparison distinguishes cumulative growth from disconnected examples.
Each later concern is assigned to a plausible later module.
The preserved law is backed by a named Module 01 learning test.

Earlier contract to preserve¶

Later modules may add boundaries and representations, but they must not erase the deterministic pure-core contract established first.

Exercise 4: Design a re-entry route¶

Starting context¶

You completed Modules 01 through 04 two months ago. You remember why pure functions matter, but you no longer trust your understanding of lazy materialization or typed stream failures. A new project pressure concerns async backpressure.

Objective¶

Design the shortest defensible route back into the course.

Constraints¶

Do not restart automatically at Module 01.
Do not jump directly to Module 08 because the project mentions async work.
Name the last contract you still trust and the first contract you need to re-prove.
Use one reading surface, one tracked state, and one focused proof before moving on.

Expected evidence¶

Produce a re-entry record:

trusted through:
uncertain contract:
first module to revisit:
source or state to inspect:
focused evidence:
move-forward condition:

Acceptance checks¶

The route revisits streaming foundations before async coordination.
The move-forward condition is observable, not “read the page.”
The route widens only after the uncertain earlier contract is trustworthy again.

Earlier contract to preserve¶

Async reasoning depends on knowing when work is produced and materialized; the recovery route must restore that Module 03 contract before adding Module 08 scheduling pressure.

Review route¶

Compare your decisions with Orientation Exercise Answers. Do not grade by identical wording. Grade by scope, ownership, preserved behavior, and the honesty of the claimed evidence.