2 min lesson
Tab troubleshooting
Complete "Tab troubleshooting". Check the final state against the lesson before you move on.
Check your understanding
Complete this walkthrough: Tab troubleshooting — fix the context before blaming the model.
Outcome: Finish with the result shown in the final walkthrough step.
Tab troubleshooting — fix the context before blaming the model
Tab in a low-context file
5 steps · ~3 min
Read the wrong ghost → Decide on the miss → Spot the pattern → State the intent → Accept the guard
Read this walkthrough as text
Tab troubleshooting — fix the context before blaming the model
Tab in a low-context file
SayYou opened searchFilter.ts to guard very short queries — and Tab proposes rewriting the return into a sort instead.
SeeConfident, plausible, wrong: nothing near the cursor says what you're here to do, so the prediction guesses at the file's shape instead of your intent.
DoReject the miss and keep going — a wrong suggestion you ignore costs nothing. (Or accept it, read what the rewrite costs, and take it back with one ⌘Z: Tab accepts are ordinary edits.)
SeeThe next ghost misses again — it flips the empty-query contract from 'return everything' to 'return nothing'. One wrong ghost is noise; two in a row is a signal about the SPOT, not the model.
DoWrite one comment above the cursor naming the goal — 'ignore 1-2 character queries' — then read the next ghost.
SeeIt's the exact guard you came to write; Tab to accept. Same file, same model — the only variable was the context you gave it. Still wrong after that? You hold a one-file repro with expected vs actual, ready to escalate.
Simulated Cursor 3.12 (macOS, light) — beta educational reconstruction, not the real product.
Learn more
Optional practice
Practice: Tab troubleshooting
QYou've applied it. What actually proves the work is done, not just a plausible answer?