The UI Era Is Over: GLM 5.2 + Codex Unlocks Token Freedom — Will the Next Generation Even Touch a Computer?
Let's talk about Zhipu's GLM 5.2, which has been taking Silicon Valley and the entire Chinese internet by storm.
I'm not exaggerating — this is our domestic ecosystem's DeepSeek R1 moment! Ever since I went all-in on this coding plan package late last year, OpenAI's pedestal and Claude's cult following have gone from sweethearts to old news. Why? Because I've achieved total Token Freedom!
What does that even mean? It means you no longer have to count every miserable penny like a broke coder with every line you write. Remember the old days? Every time you called an LLM, watching the characters scroll across the screen, your mind wasn't on how brilliant the code was — it was on "Holy crap, how much is this costing me?" Now? You can just say, "Go do all the work, and spare me the chatter!"
First, Let's Talk Token Freedom
You're probably wondering: how good does it really feel?
Let me put it this way. Back when I used Codex to debug a particularly nasty bug, the guy would just open a browser and mess around with computer use all night. Next morning, I'd wake up to find he'd burned through $200! Who can afford that? My heart was bleeding! It's like hiring a top-tier lawyer by the hour — hundreds of dollars an hour, so you never use them unless you absolutely have to. And when you do, you're watching the clock, terrified they'll say one extra word, even a greeting feels like theft. Would you dare ask them to shine your shoes or run errands? Absolutely not! Every minute, every second, you're hemorrhaging cash.
But GLM 5.2 from Zhipu completely shatters that miserable dynamic. It's like your nephew who just graduated from law school. You bring him into your firm, and not only is he cheap, but he'll do any grunt work you throw at him! Before bed, I just say, "Go refactor that product's codebase," and by morning, he's already pulled an all-nighter for an hour or two. Those long, labor-intensive tasks? He handles them flawlessly. After half a day, he's used less than 5% of his quota! Sure, he might need a couple of tries to parallel park like a new driver, but who cares? He's practically free!
So how do the big lawyer and the nephew work together?
It's a match made in heaven! The big lawyer (Codex) just sits back and supervises. When something tricky comes up — like visual recognition or extremely complex low-level logic — the lawyer steps in, gives some pointers, and stamps the final plan. Have you ever experienced that kind of unfettered, cost-free output?
You've basically got a tireless "Six Ministries" at your command! At first, I thought I'd just dabble, but soon I was completely hooked. This is the disruptive power of AI — it lets ordinary people taste the thrill of being a fat-cat boss! The grunt work is all done by tireless AI, and you just nod approval like a tyrant at the end. Who can go back to banging away at a keyboard after that?
Some people love to argue: "But GLM 5.2 has no multimodal capabilities! It can't see images! It's just a text model — a crippled product!"
Yes, it's blind visually. But have you seen the blind swordsman in Stephen Chow's movies? Blind, sure, but his swordplay is unmatched — he kills without blinking! In text processing and long-form coding, it's absurdly powerful. And precisely because it cuts out the expensive visual processing, it can be this cheap!
CLI Is the Native Language of Agents
This reveals a brutal truth about the AI world: CLI (command line) is the native language of agents!
Think about it: when agents communicate, do they need to see flashy buttons? Would two AI lawyers discussing a case send each other beautifully formatted PDFs with red annotations? No! Those green, Matrix-like lines of code are the most efficient, direct way to communicate.
You think making an AI recognize on-screen buttons and click a mouse is smart? That's stupid! That's a wheelchair for low-grade biological beings like humans! When two top-tier agents talk, a single CLI command handles handshake and data exchange in an instant. Why would they bother with visual recognition? It's like forcing two brilliant mathematicians to solve calculus problems by drawing cartoons! Their underlying jargon, executed instantly via CLI — that's real efficiency!
Graphical User Interfaces Are Doomed
Does that mean we don't need UI at all?
Have you ever asked yourself why you need UI? Do you really enjoy clicking 200 times a day in that bloated OA system just to find some stupid report buried three menus deep? Isn't that ridiculous? When we were building the Xiaoqiu Tong project, we agonized for a long time, thinking users wouldn't use it without a UI. But what happened? Nobody wants to type anymore! 80% to 90% of operations are done by tapping a single card pushed by the system. In the future, you won't even click 20 buttons a day — and even 20 feels like too many!
The UI of the future is not these fixed, zombie pages. It's called Generative UI.
Need a bar chart? It generates one instantly. Not intuitive? Switch to a line chart — it redraws in a flash. There's no fixed interface; everything is based on your real-time needs. You think you're controlling the system? Actually, AI has already fetched the information you need via CLI and databases, wrapped it in beautiful cards, and served it to you. 90% of input operations will be eliminated. The only remaining role of UI is to let you glance at the result and click "Approve" — just so you can take the blame and sign off.
The Hero Slays the Dragon, Then Becomes the Dragon
Do the bigwigs in Silicon Valley not see this coming? They see it more clearly than anyone, and that's why they're terrified. A few days ago, something absurd happened: the head of Google Workspace CLI was fired! Why?
Because he used pure command-line and AI automation to disrupt Google's bloated document, calendar, and spreadsheet ecosystem! Those arrogant executives watched their carefully built UI empire crumble — of course they panicked! Isn't this the classic story of the hero slaying the dragon, only to become the dragon? What are they afraid of? They're afraid that if nobody needs to click through those tedious interfaces anymore, what value do these giant companies that rely on monopolizing entry points and interaction surfaces have? They'd rather fire a genius than admit their era is ending. But the irony? That guy was immediately poached by OpenAI! That's the march of history!
This Transformation Runs Deeper Than We Think
So, will the next generation even need to use a computer? I'm increasingly convinced that within three to five years, most people will lose the ability to operate complex computer systems. It sounds like a terrifying cycle. A decade ago, I was the hardcore nerd who installed operating systems for every girl in my class. Today, when I get a brand new Mac, I feel overwhelmed by the complex environment setup. My first instinct? Install Codex and let AI handle all that tedious configuration! If even I can't be bothered, do you expect Gen Z to memorize those anti-human keyboard shortcuts? They don't even want to touch a keyboard!
The terminal of the future might not be a square screen at all. It could be an earring, a pair of glasses, or even just a sigh in the air. As machines begin to cater to humans, as system operations hide entirely beneath the invisible CLI, GUI — the biggest byproduct of this era — will be mercilessly swept into the trash heap of history. We're witnessing a grand, regressive evolution: it's not humans learning to use machines, but machines learning to serve humans.
Of course, if you're still feeling lost and anxious about being completely replaced by AI, I have a sincere suggestion. Instead of wasting energy on pointless worry, just buy yourself a top-of-the-line Mac Mini. What? You don't know what to do with it? Then subscribe to our People's Park Talks AI and follow along with every episode!
First, get addicted. Then, evolve.
[Subscribe to JustSayAI Daily Brief] 🌈 New friends! Subscribe to the JustSayAI Daily Brief · Two emails daily · Listen on the go 🌟 Old friends! Join our membership for the Daily Brief + in-depth reports + text columns
[Host's Xiaohongshu] 📕Call me Xiao Su
[Follow Us] ▶️YouTube｜ 📺Bilibili｜ 📕Xiaohongshu｜ 🎵Douyin｜ 📻Podcast 📱 Business cooperation/listener group: justsayai666