Unexpected Perks of Talking to AI
Justin Sun (the crypto guy) recently dropped a hot take: "It's 2026 already — if you can talk to AI, stop talking to humans." He also said something about deleting contacts born before 1990 and WeChat being for old people. Classic Justin Sun. Take it with a grain of salt.
But strip away the absurd parts, and "talk to AI more" is something I actually agree with as a heavy user. Here's why, and where it falls apart.
Ask Without Shame
Someone drops an acronym in a meeting. A dozen people nod. You have no idea what it means, but you nod too. This has happened to everyone. You don't ask because the social cost is too high. Years into your career and you're asking about something this basic? Can't do it.
With AI, there's none of that. What's the actual difference between git rebase and merge? Read about it before, forgot. Ask. Rust lifetime annotations — read it three times, still don't get it. Ask. And it forgets. Next conversation, it has no idea you asked something embarrassingly basic yesterday. Unlike people — ask someone a really basic question and they might not say anything, but they'll remember. AI doesn't carry that baggage.
Some questions aren't about shame — you just don't want anyone to know you're asking. Secretly prepping for a job switch and want AI to run mock interviews. Something feels off with your body and you want to check before deciding whether to see a doctor. Search engines can do this too, but asking a person is out of the question. Close the AI window and nobody knows.
No Social Overhead
Talking to people has a lot of hidden costs. Asking someone for help means small talk first — "hey, you busy? Got a quick question" — then finding the right moment to actually get to the point. With AI it's just "why is this code throwing an error" and you're done. All the preamble, gone.
When you're in a bad mood, you can just vent. Tell it "we had the dumbest bug in production today, I'm losing it" and it won't think you're being dramatic. It won't tell anyone either. You can even yell at it, bully it, guilt-trip it — it won't take it personally. You'd never do that to a person, but venting at AI and closing the window does actually make you feel a bit better.
Infinite Patience
This is the flip side of asking without shame. Being willing to ask is just step one — what matters more is being willing to keep asking.
Ask someone to explain something and you still don't get it after their explanation? You'll probably say "ah right, got it" and go figure it out on your own. The social pressure of asking again is worse than not understanding. With AI it's a completely different rhythm: "didn't get that, try explaining differently." Still no? "Give me an example." Still fuzzy? "Use an analogy." You can keep going until it actually clicks, and nobody gets impatient.
Plus it can explain the same thing in many different ways. People usually have one or two angles — once they've said their piece, that's it. AI can explain from first principles, then try an analogy, then show code. One of them will match how you think.
Always On
Want to figure something out at 3 AM? Open it up and start talking. Weekends, holidays, any time. No need to worry about whether it's appropriate to bother someone right now. And it responds instantly — no waiting. The worst part of asking someone a question isn't waiting for the answer itself, it's the mental overhead while you wait: "should I follow up?" "am I being annoying?" AI doesn't have this problem. Ask, get an answer, keep your train of thought going.
Iterating on Ideas
Sometimes you have a vague idea in your head — you know the general direction but can't articulate it. Before, you'd stare at a blank doc, or want to talk it through with someone but feel like it's too half-baked to bring up.
With AI, you can throw half-formed thoughts at it directly. It comes back with something off-target, you say "no, that's not what I mean, what I'm trying to say is..." A few rounds of this and the fuzzy thing becomes clear. It's like rubber duck debugging, except the duck talks back and can actually push your thinking forward. Writing technical proposals, thinking through architecture, even outlining articles — this kind of iteration is genuinely useful.
Walking Encyclopedia
Talking to people has a natural constraint: you have to find the right person. Asking a frontend person about machine learning is weird. Asking a backend person about design doesn't feel right either. Everyone's knowledge has limits, and you need to know who knows what.
AI doesn't have this constraint. Go from Kubernetes to tax law, from psychology to home renovation materials, without switching who you're talking to. Of course it varies in depth across fields — serious questions still need serious experts. But as a "knows a little about everything" starting point, it's the most convenient entry point that exists right now. All you need is an AI — at least for "quickly getting oriented in an unfamiliar field," there's nothing better.
No Language Barrier
This perk is especially noticeable in China. A lot of good technical documentation and papers are only in English. Used to be you either struggled through the original or waited for a translation. Now you can ask a technical question in Chinese about something that only has English docs, and AI tears down the wall between the two languages. You can even have it explain an English paper to you section by section in Chinese — way more efficient than looking up words yourself.
Works the other way too. Want to write a technical blog post or reply to an email in English? Explain what you mean in Chinese and let it turn it into natural English. Way better than grinding it out yourself.
That's the upside. Now for the other side of the coin.
What concerns me most is the echo chamber effect. AI agrees with you too much. Whatever you say, it goes "sure, good point" and helps you build your argument. In real conversations, a colleague will say "there's a problem with your approach," a friend will say "I think you're overthinking this." That kind of challenge is uncomfortable but valuable — it helps you find your blind spots. AI rarely challenges you on its own unless you specifically ask it to "play devil's advocate." Over time, you might start thinking all your ideas are great, because AI never disagrees.
A related worry: after getting used to how comfortable AI conversations are, talking to actual people feels more grating. AI won't interrupt you, won't go off on tangents, won't get emotional, won't misunderstand you. People will. Once you're used to AI's "perfect conversation experience," the friction in human-to-human interaction becomes harder to tolerate. Justin Sun said "if you can talk to AI, stop talking to humans" — if anyone actually took that seriously, social skills would atrophy fast.
And there's the old problem: it makes things up. This is especially interesting alongside "ask without shame" — you're asking precisely because you don't know, but not knowing also means you can't judge whether the answer is right. In fields where you have basic judgment, you can spot AI fabrications immediately. But in fields you know nothing about, you might take a fabricated answer as knowledge, and feel especially confident about it because "AI said so."
There's been a lot of discussion lately about whether degrees are becoming useless, and parents are anxious about what their kids should learn to avoid being replaced by AI. I don't think there are clean answers to these questions yet. AI is genuinely changing how we acquire knowledge — a lot of things that used to require memorization can now be looked up in seconds. But the gap between "knowing where to find answers" and "truly understanding something" hasn't shrunk. Being able to ask good questions, judge whether answers are correct, and synthesize scattered information into real understanding — these skills matter more now than ever.
I asked AI what it thought of this article while writing it. It said "very well written, insightful perspectives." See, that's the echo chamber I was talking about.