Finding Where I've Got It Wrong
Put it on the opposing side: only find errors, don't compliment
Here's my understanding of [concept]: [my explanation]. Take the opposing side and do one thing only: find what I got wrong. Requirements: 1) judge each sentence as accurate / inaccurate / vague, and for the inaccurate ones say what's wrong; 2) name the two concepts I'm conflating and distinguish them in one sentence; 3) give a concrete example where my understanding leads to a wrong conclusion; 4) don't add knowledge I didn't ask for, and don't affirm me first.
How to use
Paste the explanation you're most confident about — the errors it finds there are the ones worth having.
Good for
- You have already written your own explanation, however rough, and want the errors found before they harden
- You are about to change a production setting or put an understanding into a proposal and need someone to attack it beforehand
Not for
- Your statement is one sentence (I think it is probably a caching problem) with nothing to judge, so all it can do is nitpick wording
- You are still exploring and need a rough model that can grow, and a full correction now kills the idea
Example
Input
Concept: TCP congestion control. My statement: when TCP sees packet loss it knows the network is congested and halves the window; conversely, as long as there is no loss the network is fine.
What you get
A sentence-level verdict: the first sentence is directionally right but too strong (loss can come from link errors or reordering), the second is inaccurate (with bufferbloat there is no loss while latency is already out of control). It should name the confusion between the congestion window cwnd and the receive window rwnd. It should also give a concrete case where my reading yields a wrong conclusion: seeing zero loss, I declare the link healthy while RTT went from 20 ms to 300 ms.
Common mistakes
- Bringing two sentences you feel good about and expecting a diagnosis, when all you get is wording criticism
- Using it before you have a view of your own, which outsources the thinking and gives away your right to be wrong first
Try tweaking it like this
- Add 'pick only the two most fatal errors, one line for the rest', which keeps a review from turning into a rewrite
- Change it to 'compare my version with the textbook one, say what differs and which is better for explaining to a newcomer': correction becomes calibration