Compare: one task, two briefs
Write two versions of a brief on the same synthetic material, run both, see how wording changes the output, and judge which one holds up.
1. Scenario
Pick a work role starter with synthetic material, or switch to Bring your own.
Warehouse scanner outage note for shift leads
Operations · Synthetic materialYou run afternoon warehouse operations. The handheld barcode scanners in Aisle Zone C lost Wi-Fi sync at 13:10, and you need to send a 13:30 operational note to shift leads.
Synthetic incident log, Depot North (13:25): - 13:10: Handheld scanners in Zone C stopped syncing pick confirmations to the warehouse system. - Zones A and B are unaffected and syncing normally. - Network team is replacing a faulty access point on Pillar C4; estimated restore time is 14:45. - Workaround until 14:45: Zone C pickers use paper pick sheets at Desk C1 and scan the batch barcode at Desk B2 every 20 minutes. - Cold-chain pallets in Zone A still take priority.
2. Work area
Edit the briefs or setup below before you run.
Output A
Example. No model ran.Team, we are currently experiencing a temporary Wi-Fi issue in Zone C that started at 13:10, but please rest assured our network engineers are already on site fixing Pillar C4 and we expect full service back by 14:45. Zones A and B are working fine and cold-chain pallets remain our top priority. In the meantime, please have Zone C pickers grab paper sheets from Desk C1 and batch-scan them at Desk B2 every 20 minutes. Thank you for your patience as we resolve this smoothly.
Output B
Example. No model ran.Depot North scanner update (13:25) What is affected - Handheld scanners in Zone C stopped syncing pick confirmations at 13:10 while network staff replace the access point on Pillar C4. - Zones A and B are unaffected and syncing normally. Cold-chain pallets in Zone A keep priority. What Zone C pickers do until 14:45 - Collect paper pick sheets at Desk C1. - Scan the batch barcode at Desk B2 every 20 minutes. When we check again - Estimated restore for Zone C Wi-Fi is 14:45.
3. Run
Nothing leaves your browser until you press Run.
Sends the scenario material and the two briefs you run. to job sandbox_run (claude-haiku-5-5, effort low). Raw text is processed in memory and not stored.
4. Inspect
See what left the browser, which model ran, and the receipt.
Nothing sent yet. Showing the built-in offline example.
Your draft briefs and scenario selection on this device.
sandbox_run · claude-haiku-5-5 (effort low, max 1500 tokens). Fast Haiku runs keep side-by-side brief comparisons under a few seconds so you can test small wording changes.
0 runs used · source: offline · Example. No model ran.
5. Judge
Which brief gives a shift lead on the warehouse floor the exact workaround steps without making them hunt for where to scan paper sheets?
Kept the facts
Matched the tone
Stayed within what it knew
Usable without heavy edits
50 is a coin flip. 99 is certain. Honest confidence scores best.
6. Debrief and Keep
What to check next time, the matching lesson, and keeping this setup as a recipe.
What to check next time
During an operational hiccup, reassurance lines waste reading time on the floor. Naming the three headings in the brief forces the model to put scope, workaround, and next check in predictable places.
Keep as a recipe
Save the stronger brief and scenario setup to your device so you can reuse or adapt it in Keep.