A working example · invented event facts
Make a weekend plan you can actually use.
Three notes become a clear list. Try the result here, then give the same brief to your assistant.
The finished small tool
WRS maintainer example. It filters the supplied rows in your browser; it makes no model call.
| When | What | Where | Entry |
|---|---|---|---|
| Sat 14 Nov · 10:00–12:00 | Board-game morning | Riverside room | Free |
| Sat 14 Nov · 14:00–16:00 | Book swap | Hall B | £3 |
| Sun 15 Nov · 11:00–13:00 | Maker session | Community room | UNKNOWN |
3 events. Known entry subtotal for all three: £3, plus the maker session’s unknown price.
Copy the complete brief
You can start with a normal chat assistant. Paste these public fictional inputs and ask for the result.
What makes the answer useful?
The table preserves the two Saturday time slots, keeps Sunday separate and makes the missing price visible. The known entry subtotal is £0 + £3 = £3. That is not a complete weekend budget: travel, food and the maker price are unknown.
Now select Monday in the small tool. The correct result is an empty list with an explanation. A page that looks attractive but keeps showing Saturday has failed this specific filter check.
See a plausible mistake and the repair
Mistake: “All three events cost £3 in total.” The notes did not supply the maker price. Repair: “Known entry costs are £3; ask the organiser for the maker-session price.” This is a deliberately written teaching example, not an observed failure by a named product.
Compare two attempts fairly
Use the same brief. Keep each output and note which dates, facts or prices you had to repair. Record the actual product, plan and date. Leave time, tokens and costs unknown when you did not measure them. A correct small example is a reason to try your next similar task; it is not a general model ranking.
Make it your own
Replace the invented notes with public events you care about. Add source links and a checked date. If you want a coding project, ask your assistant to produce one local HTML page with a date filter and test the same three cases.