Learn
What is reverse engineering?
Learning how software behaves by examining the artifact itself — until you can explain a feature, change it under license, or rebuild a compatible piece.
Authorized targets only. Compliance
01 · Observe
Inputs, outputs, timing, side effects — without trusting the brochure.
02 · Recover
Locate the code path that implements the rule; keep addresses and paths.
03 · Explain
Name the rule so a teammate (or future you) can verify and rebuild.
Why agents change the loop
Evidence in, judgment stays human
A coding agent with local tools can gather strings, symbols, decompilation (via engines you own), and JS graphs faster than solo grepping. You still verify citations, refuse unauthorized targets, and gate writes.
- Ask for Evidence + unknowns — not vibes.
- Prefer absolute paths to artifacts you may inspect.
- Treat agent prose as a draft until you open the cited locations.
Good answer shape
- Named functions / addresses and file paths
- Clear unresolved edges
- No “it probably uses AI” filler
Original puzzle
Retry backoff: 1s → 2s → 4s → stop
The brochure says “smart retries.” Reverse engineering asks which function computes the delay — and what caps it.
Backoff lab
Micro-demo of the wait pattern — not a live REA session.
Grep configs and sprinkle logs.
Linear, capped, or exponential?
Max attempts, max delay, or both?
Point REA at code you may inspect; require the delay function with locations for you to verify.