Since yesterday, my WeChat has crashed several times. At first, I assumed it was just an isolated glitch, but then I noticed other friends in our group chat mentioning that both their iPhone and Mac versions of WeChat had also crashed multiple times.
Combining this with what I complained about in my friends circle a few days ago, the new black border on WeChat that allows you to press and hold to convert to text and press and hold to speak is so ugly it's unbearable. So I said in the group: What's going on with WeChat? It suddenly went from being a perfect score to a zero score? Then a friend replied: Vibe coding is harmful, I'd like to ask if you dare to live in a house designed with AI?
Although there is no evidence to prove how much vibe coding WeChat programmers use, and the recent ugly design and frequent crashes are indeed related to vibe coding. However, it seems to be a fact that WeChat is fully embracing AI. Last month, WeChat launched the AI-powered friend circle writing assistant feature, which was met with widespread criticism. Cheng Lingfeng commented, "WeChat has introduced AI-powered writing and commenting features. Disappointing. It's helping NPCs become NPCs. There are fewer and fewer real people, and the familiar relationship chain is becoming hollow. If the strongest human platform doesn't uphold the bottom line of being human, who will?" Digital life commentator Kazek also wrote an article titled "WeChat's AI New Features Are Killing the Last Bit of Human Touch in Friend Circle," which was deleted after being reported, with the complaint type showing: infringement of reputation/commercial reputation/privacy/portrait.
The last time WeChat heavily promoted Video Accounts and Mini Stores, there were already complaints that WeChat was no longer restrained. However, at the time, it didn't have much of an impact on me, because when I saw the red notification dots, I could choose not to click, or if I did click and wasn't interested, I could just exit. But this time it's different. This time, AI is polluting the core functionality of WeChat, namely the chat and Moments ecosystem. When AI-generated content dominates Moments, and in the future, if WorkBuddy is fully integrated into WeChat, with a large amount of AI content occupying private chats and groups, it will affect everyone.
In the AI era, producing something has become too easy. However, the easier it becomes, the more restraint we should exercise.
When I opened "The Product Philosophy Behind WeChat" again, it prominently stated: in 99% of cases, it's always right to reject new ideas. When you need to determine whether an idea is worth pursuing, it's simple - just reject it and don't do it.
Former TikTok user-side product manager Gill Tan said:
Xiaohongshu's MVP was a PDF that could only be downloaded one page per day, and the product's user stickiness was measured by how many people came back day after day to download it. This was a clever demand test because it used users' willingness to incur costs as a signal.
During my recent vibe coding sessions, I came across a new iteration of this logic. When you are your own target user, your MVP essentially defines how complete the final product you ship will be. Once the core functionality is up and running, the real question is whether you're willing to keep grinding through the less glamorous details—polishing the rough edges and fixing the corner cases one by one.
Because polishing details can be a painful process, if you don't really need something, your attention will start to wander. But if you're willing to see it through to a truly finished and deliverable product, that in itself is a strong signal: as a user, you really do need it.
In the AI era, the friction in product production has decreased significantly. However, the friction in matching products with users has not decreased in tandem. Therefore, just because product development has become easier, it's not possible to launch 1,000 products for testing all at once. Maintaining a certain level of friction can actually be beneficial. This brings us back to the value of signal transmission that I've been emphasizing: somewhat counterintuitively, high friction can reveal what's truly important to users.
At the start of the year, the concept of "disposable software" was all the rage, even dragging down the share prices of a number of SaaS companies. But the issue is that people still need durable goods that meet basic needs—that's the greatest common denominator of humanity. Just as daily disposable contact lenses didn't eliminate eyeglasses, and the latter's annual market size is roughly five to six times that of the former. It's like asking: when you could build a Uniqlo, why would you build a Shein? Currently, the former has a market cap of about $136 billion, while the latter is around $26 billion—the former is five times the latter.
WeChat, in my view, is a durable good that satisfies basic needs, with a value far exceeding that of disposable daily products.
It is hoped that WeChat does not lose its original intention, especially by learning from ByteDance, whose true product is its recommendation engine, rather than the superficial array of products, which later converged and became durable goods. If you use AI to create a multitude of products but do not have something like a recommendation engine to underpin them, the result will be a futile effort.
In "The Product Philosophy Behind WeChat," there's also a sentence that says: Do you want to show off your intelligence at all costs, or choose to be kind?
In my view, WeChat is a typical representative of Tencent's technological benevolence. As for AI's benevolence, I have always been reserved. AI is too smart, but it's likely that it's not inherently kind. I've said before: a large amount of text in this world is actually written by bad people, and behind the vast number of articles are various purposes, such as PR copy. So the text world has never been a true reflection of the real world. As the saying goes, "a gentleman is reserved in speech but sensitive in action," I tend to think that text reflects the worse side of the world, so large models that emerge from the text world are likely to be bad at their core.
A few days ago, I posted an article and someone commented below: "Stop writing this kind of article, it's of no benefit to yourself or others, what's the point of it?" I checked his WeChat profile and found that he's a public figure, with the title: "Member of the Democratic Party, PhD in Economics, focused on original macroeconomic analysis." He serves as the head of a Zhihu macroeconomic finance exchange group with over 45,000 members.
I suddenly realized that not everyone operates like me, an INFP, doing things spontaneously, and writing articles is often because I have to express myself and don't have a fundamental purpose.
AI always reminds me of a phrase: clever words and charming appearance, yet lacking in benevolence. I think a good product should choose kindness, choose benevolence. Obviously, AI is the opposite of this.