Just as you brainstorm 10 titles, generate 5 different thumbnail/cover concepts, and pick via a small sample, not instinct alone.

Know when to ask AI to generate new code versus review existing code — the two uses need different prompts and different caution.

Before approving any big idea, ask: what's the cheapest, fastest version testable within a week, instead of waiting for a full plan taking months?

Turn your anxious prediction into a testable hypothesis, then act and collect data.

Before any big decision, write 5 assumptions you relied on — test the weakest.

Instead of New Year resolutions, pick one "theme" (year of health, year of depth...).

Open with shocking question, surprising stat, or personal story — grab first 5 seconds.

Write your catastrophic prediction pre-event and compare with what actually happened — evidence stacks.

Mentor a younger person in new tech/culture, and learn from them in return.

Before a joke with a multicultural group, run it through a filter: does it rely on local wordplay?

Don't assume your partner's love language — test it with three situations this week and compare reactions.

Improve any process via a weekly four-stage loop: Plan, Do, Check, Act.

If you vanished for two weeks with no contact, what stops? That's your urgent delegation list.

There's no ready solution for adaptive problems — treat your next solution as a small learnable experiment, not a final plan that must succeed on the first try.

When an idea seems intuitively bad but its test cost is low, approve it as a cheap experiment instead of rejecting it on intuition alone.

For critical information, ask two different models (Claude and GPT, say) the same question independently, and compare before relying on either.

After completing AI-assisted work (report, presentation, plan), ask AI itself to attack the work as a strict external critic before submission.

For every existing operational process, ask: if we designed it today from scratch, would it look like this?