Editor-selected existing cases and prompts, separate from AI news. Selection dates use Hong Kong time.
Oct 25: five existing cases rechecked, not new or tested. Text drafts, not source agents/plugins. Gemini web lists HK/MO/TW and Traditional Chinese; ChatGPT/Claude list TW, not HK/MO. Check free/basic limits; local paid prices and Chinese quality unknown. App/API differ. Use shareable or fictional inputs and human review; no publishing, account connections, file changes or orders.
Supply a research goal, adult audience and language to draft an unsent questionnaire and flag leading wording or unnecessary personal data.
Prompt
Draft an unsent questionnaire for [research question], [adult audience] and [language]. Explain each item’s purpose, avoid combining questions, and include appropriate not-applicable or prefer-not-to-answer options. Let information needs determine length; do not default to demographic collection. Flag bias and ethics questions for the researcher. Do not publish forms, send invitations or fabricate responses.
Evidence limits and cautions
An exploratory author test, not instrument validation; this exercise produces an unsent draft.
These steps and prompts are original, untested exercises by this site, not the author’s verbatim workflow. Reported results do not guarantee exercise outcomes.
- Synthetic responses must never masquerade as real submissions.
- Real research requires separate consent, review and data-protection arrangements.
Original source: When Agentic AI Takes Charge – First impressions of Manus
Supply approved product facts, platform and length to draft alternative ad copy with evidence for every product claim.
Prompt
Using [approved product facts], draft [required quantity or range] of copy variants for [platform/audience/language], following [length and brand voice]. Explain each angle and its supporting facts; remove or flag unsupported claims. Do not add invented discounts, countdowns or testimonials. Return drafts for review only; do not activate ads.
Evidence limits and cautions
An anonymous account without performance testing; it establishes no click-through or sales improvement.
These steps and prompts are original, untested exercises by this site, not the author’s verbatim workflow. Reported results do not guarantee exercise outcomes.
- Do not invent offers, testimonials, scarcity or guarantees.
- Never put API secrets in prompts or public sheets.
Original source: Money_Invite4353 on paid-media tooling
Supply an authorized media inventory to propose title, episode, language and edition corrections without treating matching names as duplicates.
Prompt
Using [media inventory/naming rules/retention preferences], draft metadata corrections. Record work identity, season and episode, language and edition separately. Flag missing evidence; matching names do not prove duplication. Preserve original paths and show proposed differences. Report only; do not delete, move, download or upload content.
Evidence limits and cautions
A personal account does not establish that deleting apparent duplicates is safe.
These steps and prompts are original, untested exercises by this site, not the author’s verbatim workflow. Reported results do not guarantee exercise outcomes.
- Do not automatically delete, move or overwrite media.
- Do not download protected content or upload private videos.
Original source: Time-Ratio7322: home media library comment
Supply a synthetic speech transcript to draft a read-back list that tracks additions, cancellations, replacements and uncertain quantities.
Prompt
Convert [synthetic speech transcript/list fields] into a confirmation list. Track additions, cancellations, replacements and uncertain phrases. Retain units and source locations; leave missing quantities blank and ask. Provide a read-back draft only; do not connect accounts, write entries or buy anything.
Evidence limits and cautions
No comprehensive recognition evaluation is supplied; hesitation and corrections may be mistaken for firm requests.
These steps and prompts are original, untested exercises by this site, not the author’s verbatim workflow. Reported results do not guarantee exercise outcomes.
- Do not use real household recordings or private tasks.
- Do not guess quantities that were not supplied.
Original source: gargahcerk: Alexa–Todoist–ChatGPT
Supply owned recipes, target yields and fictional stock to scale and combine ingredients while flagging incompatible units and remaining quantities.
Prompt
From [owned recipes/original and target yields/fictional inventory/package specifications], build an ingredient map retaining recipe provenance, scaled amounts, combined totals and required quantities. Flag incompatible units without a supplied conversion. Calculate a list only; do not invent recipes, give health advice or operate stores.
Evidence limits and cautions
A checkout-ready list does not establish purchase completion or nutritional suitability.
These steps and prompts are original, untested exercises by this site, not the author’s verbatim workflow. Reported results do not guarantee exercise outcomes.
- Do not design therapeutic diets or promise allergy safety.
- Do not silently substitute different ingredients or package specifications.
- Hong Kong and Macao are absent from ChatGPT’s current support list. Taiwan’s listing does not guarantee the reported agent, old plugin or store is available. Analyze supplied data only; do not bypass regional restrictions.
Original source: How ChatGPT planned my meals and ordered my groceries