Foundation F07 · Lesson 06

Lab — Branches, Merge, Rebase, and Conflict Mental Model

≈ 3 min · beginner prerequisite
Objectives

By the end

  • Implement the concept without copying a full solution.
  • Compare prediction with actual runtime evidence.
  • Diagnose one intentionally introduced failure.

Lab: Branches, Merge, Rebase, and Conflict Mental Model

Purpose This lab converts the previous explanation into a small piece of evidence you created yourself. Do not begin by copying the finished example. Start with the behavior you expect, then make the computer prove or disprove your model.

Your task Create two branches that edit the same line, produce a conflict in a disposable repo, resolve it, and inspect the final graph.

Working method 1. Create the smallest clean starting point. 2. Predict the first result in writing. 3. Implement only enough to test that prediction. 4. Run it and capture the result. 5. Add one intentional failure or changed assumption. 6. Diagnose from evidence before editing randomly. 7. Restore the correct behavior. 8. Close the reference and reproduce the important part from memory.

Why this is separate from the concept lesson Reading can create familiarity without recall. The separate lab forces retrieval, execution, and debugging. Those three abilities are what later framework lessons assume when the codebase becomes larger.

Definition of done You have a runnable example, one recorded prediction, one successful result, one controlled failure with diagnosis, and a short explain-back in your own words.

Hands-on foundation studio

Predict → build → run → break → explain

Objectives

  • Use Branches, Merge, Rebase, and Conflict Mental Model in a runnable example rather than only describing it.
  • Predict a result before execution and compare prediction with evidence.
  • Introduce one safe failure and diagnose the cause systematically.

Setup

  1. Create or reuse a disposable practice folder so experiments do not pollute the production TaskFlow tree.
  2. Open the terminal and editor side-by-side. Keep browser DevTools visible for browser work.
  3. Write the expected behavior in one sentence before adding code.

Do the work

  1. 01

    Create two branches that edit the same line, produce a conflict in a disposable repo, resolve it, and inspect the final graph.

  2. 02

    Run the smallest version first. Capture exact terminal/browser/database output before adding a second concept.

  3. 03

    Change one input, ordering, state transition, or configuration assumption. Predict the new output, then run it.

  4. 04

    Explain the result in terms of runtime behavior, data flow, identity, or ownership—not “the framework did it”.

  5. 05

    Rebuild the example once without looking at the completed version.

textIndependent lab: Read the repository state before changing it
git status
git diff
git diff --staged
git log --oneline --decorate -5

# Then create one small change:
git add <file>
git commit -m "docs: explain task state"

Type it, predict it, run it, then change one assumption. The goal is not finishing the snippet; the goal is being able to explain the result.

Expected evidence

  • The example runs from a documented command or browser action.
  • Observed output matches a prediction you wrote before execution.
  • The controlled failure produces a specific, explainable symptom.
  • After the fix, the original reproduction passes again.

Intentional failure

Deliberately violate one assumption relevant to “Branches, Merge, Rebase, and Conflict Mental Model”. Keep the failure small enough that you can identify a single causal change.

Explain it back

  1. What is the simplest accurate explanation of Branches, Merge, Rebase, and Conflict Mental Model?
  2. Which input/state is the source of truth?
  3. What exact evidence distinguishes success from accidental success?
  4. What failure did you reproduce and how did you isolate it?
Checkpoint

You can continue when you can reproduce the core behavior of “Branches, Merge, Rebase, and Conflict Mental Model” from a blank file or clean state and explain one failure without reopening the finished solution.

Practice before completion

Small experiment, intentional failure, explain-back

Build the smallest example that demonstrates this lesson. Then break one relevant assumption and diagnose it using terminal/DevTools evidence. Finally explain the concept without using a framework name.