Learn Better With AI

Methods · Perspective, simulation & context

D8

Code review simulation

Have your code reviewed the way a senior engineer would review it, which is about judgement, not correctness, and is the thing courses don't teach.

When to use this
While you practise
Learning impact
Strong change
Learning evidence
Credible with caveats
Where you are with the topic
Building understanding
How often to use it
Worth repeating often
What AI is doing (1)Evaluate your work
What AI is doing (2)Play examiner, peer, or client

When to use it

  • Your code runs, but you are unsure whether a senior engineer would accept it.
  • You want practice receiving review comments without outsourcing the rewrite.
  • You keep shipping “works on my machine” solutions that are hard to maintain.

What to ask AI

Type something like: “Review this function as a senior engineer who cares about readability and failure modes. Do not rewrite it. Rank issues by damage. Wait after the first issue for my fix.”

Paste the code and state what it is for and how long it must live.

You fix; the reviewer does not take the keyboard.

How it helps

Code review as learning is apprenticeship via judgement, not a correctness oracle.

The skill is learning what seniors notice: naming, boundaries, failure modes, future readers.

If the AI rewrites everything, you practised prompting, not engineering judgement.

How to run it

  1. State purpose and expected lifetime of the code.
  2. Ask for ranked issues, not a full rewrite.
  3. Fix the highest-damage issue yourself and resend.
  4. Ask which comments were preference versus defect.
  5. Stop when you can explain each remaining trade-off.

Example exchange

You

Review this Python helper that parses dates from logs. It “works.” Don’t rewrite it,rank issues by damage and make me fix the worst first.

AI

Highest damage: silent coercion of bad dates to today. Next: the name parse_ok hides two jobs. Fix the silent coercion first, then resend.

You

Okay,now it raises on bad input. What’s still wrong?

AI

Naming still hides the dual job. Prefer a split. The defect that could corrupt downstream analysis was the silent coercion, not the name.

Copyable prompt

Review this code as a senior engineer would. Context: [what it does, who
maintains it, how long it will live]. Cover correctness, edge cases, naming,
structure, error handling and testability. For each comment, mark it [must fix],
[should fix] or [preference]. Don't rewrite it for me.

The Tell

You know this method has flipped when reviews stop finding anything because the AI rewrote your code for you.

Or when every preference is treated as a must-fix.

Keep authorship: ranked issues, you change the code.

Principle evidence

Strength of the underlying learning idea, not a claim about AI products.

The underlying learning idea is rated moderate. Code review as learning is apprenticing via feedback on authentic artifacts (cognitive apprenticeship).

AI delivery evidence

Whether an AI tutor delivers this method well is a separate question.

Principle moderate for critique-as-practice; AI senior-reviewer quality highly variable, speculative.

Related methods