Pattern Explorer
A working notebook of interface decisions. Choose a pattern, change the conditions, and test what survives.
All examples are synthetic. There are no bets, prizes, real game results or network services.
System reduced-motion preferences always take priority.
Acknowledge the action
Confirm a checkpoint with a persistent label and a changed shape.
A conceptual sequence. Timing and shapes are illustrative.
Put the pattern under pressure.
Change the conditions, then use the specimen below. These are design probes, not an accessibility certificate. Nothing is sent or saved.
Use Motion preference above for reduced motion. Narrow frame caps this specimen at 320 CSS pixels, or the available width. It does not emulate browser zoom.
Pointer activation is blocked inside the specimen and on the checkpoint command. Use Tab, Shift+Tab, Enter or Space. This switch remains available by pointer; no focus is trapped.
This local sample appears in the specimen and exported notes. Never enter personal information.
Checkpoint / local page memory
Acknowledge the action
Ready. Use Record checkpoint below.
Illustration unavailable. The written state and controls remain usable.
No stress conditions selected. The model's trade-off and failure case still apply.
State notebook 0 changes
Actual local changes, in order. Keeps the latest 30 entries; reloading clears them. No elapsed times or user measurements are invented.
Use a specimen command to record its first transition.
Checkpoint ready. Nothing has been recorded.
The sequence
- A player activates the checkpoint button once.
- The state changes from ready to recorded. A checkmark and text carry the same message.
- The confirmation stays visible. Reset is a separate, reversible action.
The trade-off
Persistent confirmation occupies space but survives missed motion or muted sound.
Where it fails
A brief glow alone disappears before a reader can inspect it. A success label must not appear before the underlying action finishes.
Three closer readings.
Use these sequences with the specimens above. The observations are about this constructed interface, not measured player behaviour.
01 / A confirmation that survives a missing picture
Select Acknowledge the action. Record the checkpoint, then turn on Missing illustration. The check-shaped artwork disappears, while the Recorded state stays written. Clear it explicitly. The log distinguishes recording from clearing; removing a picture never changes the recorded value.
Now request reduced motion and repeat. There is no suspense interval before the text becomes true. This specimen changes only memory in this page, immediately. It cannot establish that a remote save finished. In a real saving system, acknowledgement and durable completion would need separate evidence and separate states.
Trade-off: persistent feedback occupies room. It also lets someone return to the result after missing the flourish. Check that long labels do not push the reset command beyond reach. Read the full confirmation study ↗
02 / Inspect a shelf without losing your place
Select Make discovery reversible, filter to Quiet, and inspect Lantern route. Return to the results: the Quiet filter remains selected. Choose the empty filter. A message names the empty result, and Clear collection filter restores all six specimens.
Turn on Missing illustration / empty shelf. This time clearing a filter cannot restore content: the independently selected missing-content condition still applies. The distinction matters. A good reset removes the thing it promises to remove; it should not claim to repair an unrelated failure.
Trade-off: preserving context takes state. This tiny catalog has six synthetic descriptions, no real search index, ranking or persistence. Try Narrow frame and a long label together. Inspect the layout behind a narrow shelf ↗
03 / A carousel with a named return, and no timer
Select Offer a quiet way back. Advance to panel two, open its detail, then return. The second panel remains selected and focus returns to Inspect panel. Previous and Next have visible endpoints; there is no surprise wraparound and no automatic rotation.
With Keyboard only enabled, clicking a specimen command does nothing except explain the condition. Tab into the controls and activate them with Enter or Space. The switch outside the specimen is still operable by pointer, and Tab can leave the specimen. Keyboard only is a rehearsal constraint, not a global input preference or a trap.
Trade-off: buttons and the panel count take more room than an unexplained swipe. They also expose the operation before it is performed. The three panels contain original abstract studies, not commercial game covers or random outcomes. Compare the hierarchy in Keep a sense of place: that model has parent levels, while this model preserves a position on one shelf.