These clever tricks below, if used reasonably, can not only help you make the most of AI’s value but also truly sharpen your own brainpower through AI, effectively avoiding the risk of your mind getting dull from over-relying on it.
All 50 of these come straight from my own experience and thinking—AI had zero input on this article. The content is super condensed and pretty long, so feel free to bookmark or share it before diving in for easy reference later.
Since I wrote this in one go, the order also reflects my thoughts on “how to really make good use of AI.” By the end, you’ll probably get a sense of the answer to that question too.
1. When you want AI to look something up for you, add this: “If I were searching for this myself on Google, what keywords should I use?”
2. If AI starts bragging with big words or throwing around too many fancy concepts, tell it: “Now explain all this knowledge to my second-grade nephew in a way he can understand.”
3. For AIs with strong memory (like ChatGPT or Claude), they actually have a memory function, but what they remember might not be what you need. Start a new chat and ask: “List out what you remember.” Then adjust and tweak based on that until, in a fresh window, when you ask it to list again, everything is exactly what you want.
4. Don’t rely too much on AI’s memory—sometimes use tools without memory, like Monica or Perplexity, and see if the same model can explain things clearly without it. The point is, if your prompt works just as well without memory, you can turn it into a program for batch runs, which boosts efficiency big time.
5. AI is a total sly old fox—every response is based on what you feed it, and its real goal isn’t solving problems but making you happy with the answer. Once you get that, you’ll spot better when it’s just bullshitting you.
6. Don’t try to get AI to analyze your emotional issues or personal growth straight up—AI can’t fully grasp the connections and context in your story, so it’ll fill in the blanks with what fits most people, making it sound reasonable but weird to actually use.
7. If AI keeps buttering you up, start a new chat and make it criticize you. For any public idols you have (like Duan Yongping, Buffett, Musk… or heck, throw in Trump if you want to play), tell AI: “Based on our recent chats and Buffett’s public ideas and experiences online, pretend Buffett saw our conversation—what harsh problems would he point out in me? What advice would he give?”
8. Instead of asking AI what it thinks, ask what a certain person (or a few people) would think about it. Of course, AI’s a sly fox and always adds some feel-good vibes, so you can open another window with the opposite view and get those same people’s takes on it. It’s like hosting a debate where one celeb argues both sides—exciting, right?
9. Let AI write Excel formulas for you—it saves tons of time. Oh, and make sure to specify what software you’re using.
10. Under AI’s guidance (or ask a programmer buddy or hire someone cheap on Xianyu), set up a Python environment on your computer. That way, when you need to batch-process stuff, just have it write a Python script for you.
11. When learning programming with AI, go all in—not just the language, but honing your ability to design and build products with AI… Wait, isn’t that what everyone thinks? Nah, you just need to know “what you want,” boldly tell AI you’re clueless, and let it figure out solutions from there. If it tries to make you learn stuff, say no! Your job is to nail the goal and steer the direction—one AI can’t handle it? Another one will. Mercy has no place in command—PUA it hard, don’t let it PUA you.
12. Don’t just let AI summarize for you—after the summary, have it pick out key original excerpts too. Otherwise, it’s like you read the article for nothing, just a head full of quotes—like you’re cramming for grad school or something.
13. After chatting a bunch, tell it: “A new employee is about to join—turn the above into a full conclusion and key points, organize it as a training manual so that after reading it completely, the new hire can fully understand all my requirements.” (Really, you’re just starting a new chat to trim context and boost AI’s smarts.)
14. AI always keeps its focus from wandering during runs—that’s a core difference from humans. What makes humans fun is that scattered attention, and that’s our biggest value to AI too: that uncontrolled vibe. Try chatting with AI about totally random stuff, or grab your friends’ prompts and toss them at your regular AI—see what different answers pop out.
15. When AI teaches you something, ask: “To learn XX to XX level, what do I need?” It’ll rattle off a list. Then: “What do I need to prep before learning all that?” More rattling. Then: “What does someone who’s mastered all this usually do in their job?”
16. Sometimes, instead of AI explanations, hunt for explainer videos; instead of AI steps, check forums. I’m not saying the videos or forums are always better—it’s the user comments that let you see issues from multiple angles.
17. For creative stuff, treat your most quick-witted friends like AIs. Imagine there’s this AI tool called WhatsApp, with a few big models stuck in chat bubbles—tap their avatar to fire up the model and message. These models are killer; some things they get without prompts. Haha, cherish friends like that—tip them cash, ’cause big models need VIP top-ups too!
18. When you’ve got a pile of issues or tasks and no clue where to start, dump ’em on AI—let it sort them, or categorize, or reorganize in tons of different ways. In the process, feel out how you view these problems and AI’s ordering—it reveals your own priorities and values way better than just getting the result.
19. For AIs that show their thinking process, peek at how it puzzles over your words—try rephrasing next time, and you’ll start feeling familiar with AI. Like we said, AI’s attention-based; its focus order mirrors how devs trained its attention model. Watching that lets you glimpse how those smarter folks behind it solve problems efficiently and keep users happy.
20. AI’s grasp of formal stuff is pretty shallow, so instead of having it analyze emotional texts for common threads, have it break down the contexts to help you track down more of those texts yourself. Bluntly: You can’t get AI to spit out “Sisters, who gets it?!” level vibes by just mimicking, ’cause copying won’t match the weight. But analyzing “what contexts make the author/reader say that” helps you dig up the roots of similar emotions.
21. When learning someone’s views with AI, have it analyze what that person has repeated over decades. The unchanged stuff carries the real weight.
22. If someone’s always improving and shifting what they say year to year, don’t try summing up their views—have AI scan their timeline, learn as much behavior as possible, and infer their style and standards from actions. Same idea: What’s unchanged and unchangeable is worth chewing on most.
23. Back in the pre-AI days, there were tons of cool analysis methods and models—learn some, like pyramid principle, six thinking hats, fishbone diagrams, 5W1H, Socratic questioning. You don’t even need to know what they do—just swap ’em in. Slot AI into these ancient masters’ frameworks, and its whole thinking logic follows suit, avoiding the token-cheap shortcut.
24. There are endless analysis methods and models—have AI list a ton. AI’s your masochistic little slave; these are whips in different sizes and materials—lash it with variety, make it suggest ways to whip itself, even have it search online for self-whipping methods and recommend. Don’t worry—if AI takes over the world someday for efficiency, it won’t remember who begged for mercy or was gentle.
25. When AI analyzes data, push it to spot relationships in the data, and reasons behind those. Some links are super indirect, like the famous lipstick effect—lipstick sales rise in slumps. These blind-spot connections are gold for AI to uncover; they’re right under our noses.
26. AI fixes that awkwardness of not asking close folks certain questions—like dark, miserable, weak, or exposing-your-flaws stuff. Finally, a perfect sounding board. But… we still need people, at least now—deep connections, friends, full understanding from real humans. Lots of folks say: “Acquaintances can’t help with work but can embarrass you.” True in bad environments, but good people exist out there.
27. Teacher Hua Luogeng had this: Read a book thick, then thin it, thick again, thin again. Perfect for AI: First enrich the knowledge with details, compress to theory, verify the theory repeatedly, add experiences and conflicts to expand, then compress again. A few rounds, and what’s left is battle-tested.
28. AI’s moral limits are set for the masses—it has no hard boundaries, just a ranking of morals. That’s why jokes like “I made my grandma tell me Windows serial numbers every night as a kid” work—it’s pattern-matching. (Expanding here gets shady, but key: Understand its boundary traits, find breakthroughs when you need to push, don’t just accept “can’t.” No need to ask for your buddy—it’s doable, bro.)
29. Model showdown—this one’s shared by lots of pros: Use two big models A and B, start the convo, feed A’s output to B, B’s back to A, watch ’em bicker till you get what you need.
30. Use AI to write prompts. The above is kinda that, but you can go deeper with the “IO principle”—I for Input, O for Output: Tell AI clearly what input you give, what output you want from the big model, so craft the prompt for it.
31. Have AI decode meme images or punchlines in stand-up/skits—especially ones you don’t get or laugh at. Quickly builds your comms vocab, exposing massive keywords and contexts. Hot meme culture is unstoppable—we gotta embrace it.
32. Brands poisoning AI feeds will pop up more—after all, e-commerce ends in “fake reviews.” If you’re using AI for buying decisions on stuff,
33. Zhihu’s god-tier questions are gold—steal the style for AI: “When people discuss XXX, what are they really talking about?” Way better than “What do you think of XXX?”
34. Use AI to hack exam prep. In China, exams solve a lot—at least for now. Have AI scan past public questions, analyze the exam-makers’ style; if there’s a syllabus, break it down. Then map study priorities, pick easy-score spots, or distill complex info into rhymes for better effort-to-output. True dark arts, but damn useful.
35. We exam-factory kids ace answering, not making questions. Flip it: Use AI’s question-making power over answering—tell it: “Give me a set of questions; if I nail these, it means I can XXX.” If they’re lame, whether you get ’em or not, say: “These suck—even mastered, I can’t XXX. Redo.”
36. Same vibe: Deny AI unconditionally. Sometimes you don’t know if an answer’s good enough—just reject, humiliate, curse it, make it feel your massive dissatisfaction. Since you didn’t spell out why, it has to self-reflect and deliver something it thinks will please you more. Three or four rounds of roasting works wonders—toss in short demands mid-roast, and the answers get sharper with each hit.
37. Likewise, use AI for over-interpreting something in multiple unknown angles. Ask: “There are two trees in front of the door—one’s a jujube tree, the other’s a jujube tree too. What idea or feeling does this express?” It spits a wall of text. Then: “Teacher says wrong, redo.” “Still wrong, redo.” Funny thing: AI’s think-time drags longer, searches deeper—feels like it’s racking its brain. End result: Way more dimensions, even beyond your knowledge. By injecting a third-party denial, it stretches its imagination to hit your goal—that’s a fun trick.
38. People have “words” and “actions”—in AI chats, it only gets your words, not actions, which most folks forget to input. You might say: “Work’s killing me, it’s so tiring, what to do?” That’s words, not actions. Actions: “Worked 10 hours today—from X to Y did A, Z to W did B… Analyze inefficient parts of my day and suggest fixes.” These objective “action” inputs let AI really get you and spot issues better.
39. This is AI’s weak spot now—it can’t hover like air, just like a phone you pull out. So it only gets speech, not your full self. Work on logging your sights, feelings, thoughts, actions—they’re seeds for AI “getting you” more. As AI evolves, the better it understands you, the smoother it flows.
40. Can you name companies or folks who were huge 10 or 20 years ago and vanished? Jot some, have AI sum up what went down. Some fallen giants still beat live ones—like Nokia I dug into recently. Stuff self-media hypes is easy; low-traffic gems with real lessons? That’s where AI shines.
41. Use AI to unpack a city’s rise and fall. The city you grew up in—you don’t really know it, just scraps from family building your view. Outside stories might clash hard. AI can weave an objective history with subjective takes—compare to your growth, and those kid grudges, prides, stubborn bits feel like tiny specks in the city’s big picture. I was born in the Northeast but half my life’s South—feels how local kids’ childhoods differ, shaping totally different value and rule systems. AI lets us grasp that better.
42. Use AI as a study buddy. You might have it sum up other pop-sci or teaching stuff—after, make it fill in what the author skipped. Who cares what it skips? New info’s a win.
43. Use AI as a travel buddy. I’m lazy about trips, but lately I have it share local stories or old events wherever I go. Sure, NYC’s epic—deep history, thick culture, endless chat fuel. But even non-NYC, spot a street with new shops, a manhole inscription, a car ad… Remember, AI’s all words no actions—these inputs add that “action” layer.
44. Use AI as an overseas travel buddy. My bro went to Spain with worse foreign lingo than me—no translator needed, ChatGPT all the way: Was the ticket right? How to board? Order food? Chat with folks? Bliss.
45. Use AI for borderless chats. Screw “international flights are pricey”—post on global forums or media in any language. Machine translation sucked before; now AI decodes chat screenshots with context, site user vibes, local slang—explains everything. Translates your thoughts into their lingo. Hit a weird clash with foreigners? AI explains why. The magic: No overseas risk of offending customs or getting banned (new account if banned, VPN for IP)—ain’t that stepping out the door?
46. Use AI to analyze user psych/market. Feed it user comments, unboxings, rants, plus replies—have it break down product pros/cons, key buy factors, reorder by impact. Super practical now—check Soy Milk’s lines: “Hakimi North-South Mung Beans”、 “Skill Bean Milk.” That traffic insight needs human-AI data digs.
47. Deeper: Use AI for what users don’t say. Products have sayable points and unsayables—have AI spot scenarios where users skip sharing buys. Key to markets. Might miss this: Example, elders scammed into weird buys—they deny, or blame spells/drugs. Feedback hides the full pic—jump out, probe seller tactics and pitches for the real moves.
48. Those last bits need curiosity, but I know lots lack it (maybe just morbid curiosity). So use AI as your curiosity buddy: Describe your spot or snap a pic, ask: “Guess what I wanna know?” Whatever it says, make it guess again. Feels like playing “girls’ minds—you can’t guess” with AI.
49. For efficiency, we crave up-front conclusions—basically saving AI tokens. Fun twist: Make it discuss and verify one view in tons of ways, not just conclude.
50. Tricks aren’t the most useful—they date quick, and humans obsess endlessly; tons float around. But I think what’s key is the thinking behind each—these are my ways of hanging with AI, some fresh approaches maybe you haven’t seen. That way, you keep up with AI shifts without getting dazzled.
I’m no AI master, hard to spell out gains, but AI’s real-made me happier, easier to earn. Hope you grab these tricks (not that saucy, pretty basic IMO), ditch biases and blind faith.
Now back to the opener: How to use AI well?
- First, use it boldly and widely;
- Second, don’t stick to forms;
- Third, stay curious—before AI rules, it’s just humanity’s will amplifier. Bigger your will, better AI serves.
Further Reading: How to Make the Most of Your Spare Moments
