If Baidu Doesn't Build a Personal Agent Now, It's Finished

Everything I've learned in my life has been for this moment.

"Everything I've learned has been leading to this moment"

We've previously discussed StepFun's phone and the Doubao phone as the main problems with AI phones. Traditional phone makers like OPPO, vivo, and Xiaomi are now building AI phones too.

These companies have the same problem of internal interests fighting each other.

Most Android manufacturers don't make money just from selling phones — a big chunk of their revenue comes from pre-installed software, ad fees, and revenue sharing from users' subsequent use of those apps.

Do you really think Lei Jun sells a one-or-two-thousand-yuan Redmi just to make friends with you?

Sure, he wants to make friends, but mainly he wants you to spend more time on that Redmi looking at splash-screen ads and downloading apps from the "Today's Picks" folder to watch more ads.

Suppose OPPO actually built a really great AI phone today, with a fully open ecosystem and an Agent that can call everything for you. Those users would then stop using apps and stop looking at splash-screen ads — and OPPO wouldn't earn ad money anymore.

There's a very real chance that OPPO's AI phones sell better and better while short-term revenue actually falls.

And it's hard to charge users for these AI assistants themselves. Token rates are something Chinese people have a hard time accepting.

Buy a phone, and you not only pay for data — you also pay for tokens. In that case, I'd rather just turn it on and watch ads.

Of course, Apple isn't any better — the Apple tax is pure gangster logic.

In summary, it's very hard for traditional Chinese phone makers to change their old ad-selling business model, and selling ads is inherently at odds with the AI phone's core purpose of boosting efficiency.

But don't lose heart. I've found the most underrated company that actually has the best shot at building a PA and an AI phone — Baidu.

Hear me out.

From the discussion above, we already know that for AI phones, beyond the hardware itself and Agent capabilities, the most important thing is the software.

Alibaba, ByteDance, and Tencent each block each other's ecosystem positions — none would let a rival plug into them — but that's secondary.

More importantly, a chunk of these traditional players' revenue (Taobao, Ant Group, Meituan, etc.) comes from the B2B side, which needs DAU to sell B2B ads. They won't be willing to let users just call things through an Agent.

For example, even if Ant Group were willing to grant payment permissions, the Alipay department would never agree.

Otherwise, why do you think Alipay makes short videos and Ant Forest? Isn't it just to keep you inside Alipay longer?

The Taobao coupon team would never let Taobao AI redeem their coupons properly — otherwise the coupon team could pack up and leave.

Failure within your sphere of influence is unacceptable; success outside it is unsustainable. These traditional giants will fight until AGI arrives and still won't open up the walls between departments.

A friend of mine suggested the business model could shift to selling GEO rankings. Maybe that works, but self-revolution is always hard.

So the only way to break the deadlock is for a new player to enter.

And right now, I think of Baidu.

When we talk about Personal Agents and AI phones day to day, it always comes down to a few scenarios: first, food delivery. Second, shopping. Third, ride-hailing. Fourth, booking flights and hotels.

Hey — Baidu has done all of these, because Baidu has been through every internet business it was supposed to do: Baidu Waimai, Baidu Nuomi Films, Baidu Travel, Baidu Hui Life, Baidu Mall...

Food delivery, shopping, flights and hotels — it mastered them all.

As for ride-hailing, Baidu really never did that, but Baidu has Apollo Go — no driver needed. Open your AI phone, give a shout, and a robotaxi pulls up. Isn't that straight out of sci-fi?

These businesses can also work with Baidu Maps to point the way to Baidu's future.

Having done them isn't the most important thing. The most important thing is that these businesses won't run into the internal competition you find inside traditional apps like Alipay or Meituan — because there's nobody left in these businesses.

Dig the code for these apps back out, onboard the merchants, serve consumers purely, don't chase ad money, and scrap the coupon team — just hand out perks to the family, callable only by your own AI phone.

Case closed, folks. You've been dunking on Baidu all these years, but Robin Li was actually thinking further ahead than all of us 😭

Over the past decade-plus, Baidu has done every kind of business — and conveniently, it failed at all of them. Only because they failed is there no internal resistance. That's what "everything I've learned has been leading to this moment" means.

Baidu even has Du Xiaoman. Imagine: you tell your AI phone to buy you an iPhone, and if you're short on cash, it just swings by Du Xiaoman and takes out a loan for you. How slick is that.

Why would Baidu even bother with Baidu Dazi? People might think of SenseTime's Xiaohuanxiong, but they'd never think of Baidu.

But if Baidu jumps out today and says it's making an AI phone, that's a masterstroke. ByteDance, Alibaba, Tencent — you keep grinding on models, keep grinding on office Agents. Baidu will just slip in through the back door.

I've heard "the winner in AI is WeChat" for three straight years now — this hope that WeChat will take over everything, this hope that Zhang Xiaolong will unleash his ultimate move. That's a very boomer mindset.

Today we put forward a novel take: Baidu is the winner in AI phones!

And Baidu doesn't even need to build apps now — just ship a CLI and let Agents call it.

I see Ant Group's Lingguang and Today AI all buying ads to hype themselves as the Chinese Muse. Nonsense. The Chinese Muse is Baidu — resurrecting all its failed businesses from the past!

(This article's cover image was generated by ChatGPT. Purely human-written.)

⬇️

Welcome to subscribe to our Substack: funeralai.substack.com