Doubao Phone Is a Useless Pain for Doubao

"IP Merch" Yesterday Qiu Mu called the StepFun phone the "surrender faction" because it took the A2A route, which meant it could only call services if rival companies opened their APIs.

I don't buy that argument, because by that logic, wouldn't the Doubao phone be the "resistance faction" for taking the GUI route? But the Doubao phone is pretty useless too — so aren't we just giving it too much credit?

Before this episode of Idol TV starts, I must first apologize to Doubao. I previously described the Doubao phone as incapable of doing anything — couldn't send Lark messages, couldn't order takeout, couldn't book hotels.

I was wrong. The Doubao phone can actually do everything.

Because the Doubao phone is integrated with Douyin, and Douyin has takeout, hotels, and other services. So if you're willing to spend the time directing Doubao to work, it can technically get things done.

Douyin actually has takeout. Chain stores like McDonald's and Luckin Coffee have mini-programs on Douyin with built-in delivery. You buy a group voucher on Douyin, then jump to the McDonald's mini-program, select immediate delivery — and yes, the Doubao phone can walk through this whole process for you.

It's just that GUI is slow, Douyin has pop-ups everywhere, the coupons are a mess, and Doubao keeps asking me to take over. The first time the Doubao phone completed this flow, it took 18 minutes and 34 seconds. That doesn't even count the two times I had to intervene mid-process.

So yeah, the Doubao phone being able to do everything and being useless aren't contradictory. It can do the work, but there's no reason to make Doubao do it.

And this is after a month of optimizations. Go back to a month ago, right after launch, and I was actually kind of angry when I first got the Doubao phone.

Because in those first few days, the AI features basically didn't work at all — nearly shattering my stereotype of ByteDance.

My long-standing stereotype has been: big tech companies aren't very innovative with products. ByteDance isn't very innovative either, but at least their stuff is always smooth and polished.

I spent a hefty 5,999 RMB on the base model. Not being able to call Tencent or Alibaba apps was expected. But when I asked it to send a message to a Lark contact and it failed — that was testing my patience.

A couple weeks ago, the Doubao phone experience looked like this ⬇️

My assessment: too many bugs.

Of course, I believe in Doubao. I believe the gravity of mediocrity hasn't fully enveloped the Doubao planet yet.

To show respect for Doubao, I used this phone for over a month and even took it to the Strait of Hormuz to see the sunken American ship.

Unfortunately, ByteDance products don't work in Iran at all. I used the Doubao phone there by opening Qwen in a floating window, taking screenshots and asking Qwen directly — basically the same as using the Doubao phone normally. So for a while I thought the Doubao phone should be renamed the Qwen phone.

But these past few days, using the Doubao phone again — damn, three days apart and I'm impressed. They've got the basics working.

First off, the Doubao phone can actually use Lark properly.

The AI workflow goes like this: you say "Doubao" twice, the input box pops up automatically, and you can talk directly. There's also a physical AI button on the phone that wakes up Doubao with one press.

I told it to send a message to Qiu Mu, create a Lark meeting for tomorrow, dictated a long passage and had it write up a Lark doc to send to Qiu Mu. All these Lark tasks, it can handle relatively quickly.

This isn't exactly revolutionary — Lark and DingTalk both opened their CLIs, practically begging agents to integrate and build out the office ecosystem. Failing to call Lark would be an extreme failure.

But an AI phone being this corporate is kind of embarrassing.

No choice though. After getting banned by other apps one after another, the Doubao phone really has nothing to do with entertainment. I asked it to scroll Douyin Lite for rewards, but Doubao refused after five minutes, saying it involved sensitive operations and pausing.

So besides very corporate office scenarios, can the Doubao phone do the agent features phone makers love to hype — ordering takeout, booking hotels, shopping?

This is where I owe Doubao a serious apology.

I initially thought these were impossible for the Doubao phone, because basically every major app you can think of refuses to be called by Doubao — Alibaba, Tencent, Meituan, Kuaishou, plus Xiaohongshu, DiDi, Pinduoduo, and 155 apps total.

Just yesterday, 22 more apps joined this list. Now even HUPU won't let Doubao call it. HUPU is really keeping business and pleasure separate here, because last week HUPU users just voted Doubao as HUPU Goddess.

At this rate, even if these major apps open up to various terminals in the future, they'll very likely block the Doubao phone specifically. After all, ByteDance's ambition is massive — there's nothing ByteDance wouldn't do, and everyone's scared.

But this is exactly where ByteDance is brilliant: Douyin has achieved a closed ecosystem loop, with full local life services.

And this is exactly where the Doubao phone's GUI route shines. Takeout is a neglected corner of Douyin's business — normal people don't order takeout on Douyin, they buy group vouchers.

But if you insist on having the Doubao phone order takeout, it taps and taps and taps, and yes, it can order takeout.

Same for e-commerce and hotels. The experience is terrible though, because the Doubao phone is too safe — it absolutely refuses to tap pay, and even permission confirmation pop-ups sometimes require me to take over.

The interaction method is also awful. I asked it to pick a hotel, and it sent me a wall of text in the chat box. I had to go back to Douyin and search each hotel one by one.

But it is technically useful. Because the Doubao phone can also call Trip.com Group for price comparison, I discovered that Douyin hotels are way cheaper. Same Hangzhou hotel: Trip.com Group lists it at 700+ RMB, Douyin voucher is just 500+ RMB.

But that's about it. Can't say it's useless, but it's not exactly useful either.

The second-gen Doubao phone's marketing特别强调 this isn't the Doubao phone, but rather the "nubia NaviX Ultra with Doubao Phone Assistant Consumer Edition, co-developed by Doubao and nubia."

So setting aside the AI component, evaluating it purely as a nubia phone — it's just a standard Android flagship.

My first impression unboxing the nubia NaviX Ultra was better than expected. The phone looks mediocre in photos but feels decent in hand — matte glass back, aluminum frame, with "POWERD BY DOUBAO" engraved in small letters up top.

Pure text doesn't convey this, folks.

All I can say is, my daily driver is the iPhone 17, and the nubia NaviX Ultra's build quality is no worse than iPhone's. The camera module is one big horizontal slab that's very protruding, but still better than the massive circular camera slabs that were trendy on Android phones a few years back.

But here's the thing: if the Doubao features are mediocre, why would I spend 5,999 RMB on a nubia? 😭

Android phone competition has reached absolutely insane levels. Every manufacturer is going mad with aggressive innovations. HONOR put a gimbal in their flagship, Xiaomi made a rearview mirror, OPPO even has Manus now — who knows if Manus is going to replicate the autonomous driving company ecosystem niche.

At the same price point as the Doubao phone, Huawei's wide bar phone — I tried it once at a Huawei flagship store in a mall — comprehensively destroys the Doubao phone in build quality and innovation.

If you can't innovate on AI features, can't you at least do something new with the hardware? Something like Huawei's wide bar form factor would be great — way more comfortable to hold one-handed than a long slab.

You can't fully blame nubia for the hardware form factor either. As the client, how could ByteDance not channel some geek spirit and make a cool wide bar phone to show off?

As for photography, the Doubao phone is purely nubia photography — that overexposed, oversaturated Android photo quality, comprehensively behind Huawei, OPPO and other Android flagships, and not as good as Apple either.

The background is the beautiful Strait of Hormuz. The Doubao phone's photos deviate quite a bit from what the eye sees — way too vibrant.

The Doubao phone has three cameras: 0.7x, 1x, and 3x zoom. But when you switch focal lengths, the frame noticeably stutters.

I don't know how nubia managed this. My 3,000 RMB OPPO from last year doesn't do this when switching lenses.

I ran into someone from a phone manufacturer recently. They played with the Doubao phone for a bit, then excitedly pulled out their own phone and started recording the zoom stutter, saying they were taking it back to show their colleagues for a laugh.

This thing shooting the beautiful Homa Hotel lobby in Bandar Abbas can still overexpose. Apple photos are a bit dark out of camera, but tap once for auto-adjust and it looks fine.

I'm serious, the first-gen Doubao phone could be excused as circumstances beyond its control. The second-gen is genuinely damaging Doubao's public image.

Mr. Commander-in-Chief Zhao, can you bear to see Doubao's good name tarnished? Mr. Supreme Commander, can you bear to see the gravity of mediocrity drag Doubao down?

Do it, sir!

Phones are a fully mature, hyper-competitive industry. The mainstream manufacturers that have survived all have real chops, real product innovation capabilities — not AI teams that rode a hype wave and think they can compete.

For all AI phones planning to launch domestically, I have two concrete suggestions.

For AI phones priced at 6,000 RMB and above, you'd better have WeChat integration, or it's definitely not worth the price.

Of course, since you probably can't win over Brother Long, you'd better do something novel instead. No rule says AI phones have to be full-screen, right?

Reviving keypad phones — I see real potential there, at least for collectible and ornamental value. Worst case, figure out how to add some AI features to Nokia, use it purely as a remote control connected to your own agent?

I think the Doubao phone is perfect for this approach.

One-sentence selling point: the Doubao phone is a Lark remote control, the ultimate work phone for ByteDance employees.

Considering ByteDance has over 100,000 employees, if everyone gets one Doubao phone, a pure ByteDance internal loop could probably sustain a small phone ecosystem. Hand them out as Spring Festival employee benefits, learn from NVIDIA and model factory best practices, create circular sales to drive revenue flywheel.

Of course, I'm still fully confident in the Doubao phone.

First because I'm absolutely loyal to Doubao. Second, I believe the Doubao phone will figure out Douyin Group's various services — takeout, hotels, restaurant reservations — way faster than other manufacturers waiting for external ecosystem openness.

While everyone's discussing whether other ecosystems will open up, don't forget Douyin has everything. When the time comes, the Doubao phone GUI-ing Douyin's full local life suite will destroy you all.

My other thought: AI phones and personal agents are fundamentally the same thing — just a matter of whether you approach from software or hardware.

These scenarios AI phones have been recycling for two years — having AI run errands for you, order takeout, book hotels, make calls — too convoluted. I stare at the Doubao phone input box and my mind goes blank, can't think of anything I need it to do.

The actual tedious tasks I need AI help with — processing invoices, paying utilities, organizing visa documents, filling out various forms — these AI phones can't handle because they don't have WeChat integration, no context, not smart enough, can't keep a browser running long-term.

This is the story personal agents tell: you need an agent running in the cloud, connected to all the user's devices, accessing all context, analyzing user intent, scheduling tools, environments, and swarms of mini-agents to complete tasks based on the user's offhand remarks.

Users can find this agent through the phone system, SMS, calls, any entry point — accompanying them 24/7.

However, Junyuan Qi probably isn't doing the StepFun phone anymore, and who knows what the Doubao phone can do — maybe have Manus acquire OPPO?

Use beautifully Shenzhen-made phones to revive the great general agent, and let that mess of personal agents go to hell!

(Cover image generated by ChatGPT, purely human-written)

⬇️

Subscribe to our Substack: funeralai.substack.com