• The standard task-switching paradigm: one card, two questions, the colour says which
  • Both runs hold identical work — same cards, same number of each question, proved before either is played (npm run verify:switchcost)
  • The blocked run goes first on purpose: practice favours the second run, so a gap that survives it is real
  • No verdict on the person — it closes on the move that works: batching
  • Arrow keys or F/J work, because this one is measured in milliseconds
  • Keep your own numbers across daysplanned

✓ = done · ☐ = planned. Every game is built one step at a time like this.