Daily picks

Rain threatens your trip? Research alternatives with ChatGPT

Use ChatGPT search with your dates, base and walking limit to draft sourced indoor and transport alternatives; check opening information yourself.

Prompt

Use web search to plan a one-day walking itinerary in [city] on [date], starting from [area], between [start and end time]. Limit each walking leg to [minutes]. Must-visit places, breaks and other needs: [requirements, or none]. Provide sunny and rainy versions, transport options, reservation requirements and official sources. Mark unverified opening hours and estimated travel times clearly. Do not book anything.

Evidence limits and cautions

• Prices, schedules and opening hours can change; verify them again.
• Estimated journey times are not live navigation. Leave a buffer.

Original source: OpenAI：Searching the web with ChatGPT

Need another way to review? Make a NotebookLM audio overview

Add authorized study material, set the review focus in NotebookLM and generate an audio overview; return to the source for unclear points.

Prompt

Create an Audio Overview grounded in the selected sources, focusing on [chapter/topic/concepts]. The audience’s prior knowledge is [level] and their learning goal is [goal]. Explain core concepts at that level. If using analogies, clearly separate them from source facts; do not add concepts to meet a quota. Finish with questions that still require checking the original material.

Evidence limits and cautions

• Generated audio can omit details or make mistakes; it does not replace the original.
• Language, length and options depend on your current account interface.

Original source: Google：Customizing your Audio Overviews

Function hard to maintain? Ask GitHub Copilot for a small refactor

Run baseline tests on a branch, specify one function and behavior to preserve, then review Copilot’s small refactoring proposal and rerun tests.

Prompt

Propose the smallest refactor of [target function/code section] to address [problem to improve]. Preserve existing behavior, public interfaces, error handling and output format; add no dependencies. Explain scope, risks and verification steps. Flag unclear requirements or missing tests, then wait for my confirmation before editing.

Evidence limits and cautions

• Add tests first when coverage is weak; do not assume behavior is preserved.
• Do not automatically merge, deploy or upgrade dependencies.

Original source: GitHub Docs：Refactor code

Draft social posts with a fact-check list

Using Claude where available: Supply an original article, audience and length limits to get draft posts with sources and open checks.

Prompt

Adapt [owned original] into draft posts for [platform, audience, tone and length]. First list supported facts and sources and distinguish opinions. Never fill gaps with invented numbers. Keep review notes, do not schedule or publish, and do not force a controversial angle.

Evidence limits and cautions

The failed agent’s platform is unspecified; the aggregator’s Zapier classification is not adopted.

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 controversy or statistics for attention.
• Tool choice needs separate verification; this does not establish a particular platform’s failure.

Original source: Zapier CEO Just Revealed What Most People Get Wrong About AI Agents

Too many chores? Make a plan that fits your time

Supply chores, fixed commitments and available time to draft a flexible plan with priorities, buffers and explicit tradeoffs.

Prompt

Using [routine chores, fixed commitments, available time and preferred detail level], draft an adjustable chore plan. Distinguish necessary, deferrable and unclear work, showing estimates and buffers. If capacity is insufficient, explain tradeoffs rather than overfilling the schedule. Do not assign family responsibilities by assumption, plan hazardous repairs, create reminders or modify calendars.

Evidence limits and cautions

This is subjective experience from one user, without measured completion rates or health outcomes. Only a relative date is visible.

The following is an original, untested exercise, not a reproduction of the reported session. Check available features and permissions, then adapt it to your own constraints.

• Do not inherit the source’s minute limits or household size. AI does not know the home’s actual condition.
• This is ordinary planning, not mental-health care or professional equipment maintenance.

Original source: Using ChatGPT to schedule my day, remember my tasks, and help me motivate to do them has literally changed my life