Explain Complex Code in Simple Terms

This prompt turns an unfamiliar block of code into a layered explanation aimed at a junior developer, moving from a one-line summary down to step-by-step logic, tricky patterns, and possible bugs. Reach for it when you inherit a file, review a teammate's pull request, or land in a codebase whose conventions you don't yet know. The four numbered sections force the model to separate 'what it does' from 'what's risky', so you get both comprehension and a code-review signal in one pass.

What you can expect back

1) Summary: This is a memoize decorator — it wraps a function so repeated calls with the same arguments return a cached result instead of recomputing.

2) Step-by-step:
- `cache = {}` creates a dictionary that lives as long as the decorated function does (a closure).
- `wrapper(*args)` accepts any positional arguments.
- If `args` hasn't been seen, it calls the real function and stores the result keyed by `args`.
- Otherwise it returns the stored value. `return wrapper` swaps the original function for this caching version.

3) Tricky bits:
- It relies on `args` (a tuple) being hashable, which is why it can be a dict key.
- The cache is shared across all calls and never cleared.

4) Bugs / improvements:
- It ignores keyword arguments — `f(1)` and `f(x=1)` won't share a cache entry and kwargs will actually raise.
- The cache grows forever (memory leak risk); consider `functools.lru_cache` with a maxsize.
- Add `functools.wraps(fn)` to preserve the original function's name and docstring.

Illustrative example — your results will vary by tool and inputs.

Get sharper results

• 01Paste the surrounding imports and any called helper functions too — the model can only judge 'potential bugs' against the context it can see.
• 02If a specific line is confusing you, add a one-line note like 'I especially don't get the closure part' so section 2 zooms in there.
• 03For a stricter review, append 'Rank the bugs in section 4 by severity and tell me which one to fix first.'
• 04When the language is unusual or domain-specific (Verilog, COBOL, a custom DSL), name it explicitly so the model doesn't misread the syntax.
• 05Ask it to rewrite the function with its suggested improvements applied as a follow-up, so you can diff the before and after.

Adapt it for your case