AI News

Luo Yonghao sells TRAE in Double 11 livestream: the 9.9 yuan AI Coding product is a hit

36Kr · AIUpdated Nov 11 18:10

AI summary

In his Double 11 livestream, Luo Yonghao spent less than 60 minutes demonstrating TRAE, ByteDance's AI development product, and put the Lite membership on sale at a first-month discount of 9.9 yuan. On the spot, he used a single sketch to build "Mantou Voice," a voice input app triggered by the mouse wheel, and also had TRAE generate an in-depth report on the AI short drama market and two sets of Keynotes. As of June 2026, TRAE has more than 10 million registered users, covering more than 200 countries and regions.

Body · AI translation

How fast is the era changing? The Double 11 livestream rooms are the first to know.

No sooner had Li Jiaqi sent Token trending on social media than Luo Yonghao arrived with AI Coding.

In the 2026 Double 11, top livestreamers all started selling AI.

If Token is still the raw material of the AI era, then AI Coding is more like the commodity of the AI era. For ordinary people, the shift from Token to AI Coding may be the distance from buying AI traffic to buying AI services.

A livestream of less than 60 minutes turned AI Coding from a complex developer tool into a "new productivity consumer product" that anyone can buy and anyone can pick up.

The AI Coding tool TRAE, which went on sale in Luo Yonghao's livestream room, is also becoming a focus of public attention.

This Double 11, Luo Yonghao sold TRAE like crazy

This Double 11, a rather peculiar scene appeared in Luo Yonghao's livestream room.

While other livestreamers were busy talking prices and urging orders, Luo Yonghao was facing the computer screen, enthusiastically showing the audience how he developed software. Right in the middle of his excitement, a colleague beside him couldn't help but interject, reminding the audience that the 9.9 yuan purchase link had already been listed, only to be cut off by him:

"Don't talk about that commercial stuff first—you're making my lesson vulgar."

After all, what Luo Yonghao was selling this time was not an off-the-shelf product, but an AI tool that can help ordinary people create software themselves. To prove this to the tens of thousands of viewers in the livestream room, he even spent several hours in advance personally acting as a software developer.

Back when he ran Smartisan Technology, Luo Yonghao needed to find product managers, designers, and engineers to develop software; this time, he handed all that communication to an AI engineer—TRAE.

He even gave this name a Chinese-style nickname: "Cui Laosan."

At 5:49 p.m. on November 10, Luo Yonghao opened TRAE, ByteDance's AI development product, and entered his first requirement:

"I want to make a voice input method triggered by the mouse scroll wheel."

To make it easier to explain, he drew a sketch himself.

TRAE not only understood the sketch, but also proactively asked about feature details, usage scenarios, and compatibility issues. What moved Luo Yonghao most was that in the past, when communicating with engineers, if he suddenly changed requirements halfway through, he would inevitably get a "Why didn't you say so earlier?"

Now, facing this AI engineer he affectionately calls "Cui Laosan," Luo Yonghao can finally act confidently as the client: "Even if I change it 100 times, he doesn't complain."

Of course, a beginner's development process can't be smooth sailing all the way.

To cover up his mistake, he was even too embarrassed to tell the AI that he had chosen the wrong model, and instead falsely claimed he needed to switch to another computer, asking TRAE to package all the completed code, project materials, and subsequent development instructions so he could switch models and continue.

Just like that, an idea that was still in Luo Yonghao's head in the evening really became a runnable piece of software late at night.

Luo Yonghao named it "Mantou Voice." The functions of Mantou Voice are actually very simple:

Turn the mouse scroll wheel into a voice input button that can be activated at any time. When you want to type, you don't need to switch input methods—just move the mouse to the text box, hold down the middle scroll wheel button, and start speaking, and the text will automatically appear at the cursor position. To add punctuation, move the mouse to the corresponding button and release; if you say something wrong, you can delete it directly; in chat windows such as Feishu and WeChat, you can also choose to send directly.

At the livestream, Luo Yonghao even used it to send a message to a colleague:

"Let's go have a couple of drinks after work."

Of course, this freshly made software is not yet perfect. During the demonstration, it occasionally swallowed a character, and the send function also temporarily failed. But this was the first software project he completed independently with the help of AI, without looking for designers, product managers, and engineers.

A person who doesn't know how to program, relying on a sketch, a few rounds of dialogue, and constantly revised requirements, really built the app he wanted.

In addition to developing software, he also demonstrated two capabilities that may be even more appealing to ordinary office workers:

Doing in-depth research, and making PPTs.

Friends who often work know that on weekends, office workers most fear their boss suddenly assigning an industry report due on Monday. Collecting materials, verifying data, making charts, communicating revisions... even sacrificing two precious weekend days may not be enough to get it done.

Luo Yonghao is clearly also well versed in workplace ways. He asked TRAE to produce an in-depth analysis report on China's AI short drama market from January to October 2026.

According to the records shown in the livestream, the first version of the report took only 3 minutes and 57 seconds in total. But Luo Yonghao was not satisfied. He asked TRAE to import the report into Feishu, use pie charts wherever pie charts could be used and bar charts wherever bar charts could be used, to appear as earnest and attentive as possible, and instructed TRAE:

"You have to fool the boss, but you can't deceive the boss."

TRAE then called Feishu tools, created a document, and added charts. Midway through, the computer unexpectedly crashed, and after it recovered, the AI continued and finished the work.

Luo Yonghao simply had TRAE stand in the boss's shoes, predict the most likely questions, and organize the answers and supporting data into a list in advance. Looking at the generated "boss briefing response checklist," he was quite satisfied: "It predicted the boss's predictions."

As for another nightmare of office workers—PPT—Luo Yonghao likewise handed it to TRAE.

He uploaded a research document of more than 50,000 words and asked for two sets of Keynote presentations: one for speaking, with clean pages and detailed content placed in the speaker notes; the other for sending to others to read, understandable without listening to a narration.

TRAE suggested 16:9, but he thought it wasn't wide enough and changed it to 21:9; he also specifically asked TRAE to work in the background, so it wouldn't occupy the screen and delay his own work.

Mouse input methods, in-depth industry reports, presentation PPTs—these tasks look unrelated, but they actually point to the same kind of change:

AI is not only responsible for "Chat" but has also begun to execute.

Understanding goals, planning steps, calling tools, checking results, and continuing to revise when problems arise... this is the value of the Agent-style workflow.

And TRAE, whose full name is The Real AI Engineer, meaning "the real AI engineer," is precisely an AI development product based on this way of working.

What also makes this kind of attempt easier to happen is TRAE's pricing.

During the livestream, Luo Yonghao compared the frequently appearing Tokens to mobile phone data: simple tasks consume little, while complex development and multiple rounds of revisions consume more.

TRAE converts resource consumption into credits and offers three membership tiers—Lite, Pro, and Ultra—corresponding to different usage needs.

This year's Double 11, the Lite membership first-month discount for first-time paying users is only 9.9 yuan.

Luo Yonghao encouraged the audience more than once to spend 9.9 yuan first and try it out for a month themselves.

After all, in the past, if you wanted to develop a piece of software, even just a simple little tool, you might have to pay professionals to help.

Now, at least, you can start trying AI Coding for the price of a cup of coffee.

AI Coding is becoming a "mass consumer product"

For people familiar with AI programming, TRAE is already nothing new.

According to the introduction in Luo Yonghao's livestream, as of June 2026, TRAE had more than 10 million registered users, covering more than 200 countries and regions;

users used it nearly 5 days a week on average—in other words, "using it every workday."

Luo Yonghao regarded this as something to be alert to:

some people have long been learning new tools every workday, while others still only know to occasionally open a chat box and ask a couple of questions.

If the past 20 years were about trying to close the digital divide, then in the future, the new gap may come from the "AI divide."

Luo Yonghao also remarked during the livestream that Chinese consumers have historically been accustomed to paying for hardware such as phones and computers, but are less willing to spend money on software.

Now, we can use AI Coding to turn professional experience into tools, and let the tools in turn help us work.

Sometimes, the best product ideas do not come from the tech world either.

As one of the earliest companies in China to work on AI Coding, TRAE has, through official hackathons, Meetups, TRAE Friends, TRAE on Campus, and other community events, reached more than 60 cities worldwide, held more than 130 offline events, and attracted more than 20,000 developers to participate.

But in the livestream, Luo Yonghao shared several cases of ordinary people using TRAE to participate in an AI creativity competition.

Two high school students who like badminton, because it was difficult to see their own swing movements clearly and hiring professional coaches over the long term was too expensive, tried to use TRAE to develop an app that analyzes swing movements.

A young person did not want elderly family members to live under cameras for safety over the long term and lose their privacy, so they tried to develop a reminder tool that judges abnormal situations through a pillbox and door sensors.

There is no need to film the elderly person's life all day, and family members can still be reminded when medication is missed or abnormal comings and goings occur.

According to Luo Yonghao, these participants originally did not know how to write code.

Sometimes the most valuable product ideas come precisely from specific problems ordinary people encounter in life.

A person may not understand code, yet may know better than a programmer what kind of tool they actually need.

In the past, such ideas often stopped because of a lack of technology, funding, or a team.

Today, AI Coding is offering another possibility: first build the idea, then gradually improve it.

Of course, generating code does not equal delivering a product.

Programs may report errors, features may fail, and what really gives beginners headaches is often not knowing how to fix things after a problem arises.

This is also why AI development platforms such as TRAE are worth paying attention to.

In Agent mode, we can start from a goal described in natural language and let AI help break down requirements, execute tasks, and advance the project.

When there is a need to inspect the program in depth and adjust details, we can switch to the IDE development environment for more fine-grained viewing, modification, and debugging of the code.

It is worth noting that TRAE does not mean human review can be completely skipped.

Projects involving real users, sensitive data, expenses, and formal launch still require sufficient testing, security checks, and necessary professional review.

And when a person can not only make tools for themselves but also try to turn tools into products, the significance of AI Coding takes another step forward.

In 2026, "One Person Company" (OPC) became a hot concept in the entrepreneurship field.

It no longer merely refers to a company with only one shareholder in the legal sense, but represents a new entrepreneurial possibility:

In the past, when we had a startup idea, even if we just wanted to develop a simple piece of software, we needed to find programmers, designers, or even form a team.

Before the idea could be validated, the money was almost burned through.

Now, with TRAE, one person also has the chance to build a product prototype from scratch, validate ideas at lower cost, and even gradually build their own business.

The first step in starting a business may no longer be finding people, finding money, and forming a team; it can also be opening a computer and first building out your own idea.

Of course, a lower technical threshold does not mean entrepreneurship will necessarily succeed.

Whether the market is willing to pay and whether the product can continue to operate are still difficult issues all entrepreneurs must face.

But at least, TRAE is expanding the boundaries of what one person can do.

For ordinary people like us, we can start with the problems we are most familiar with, use TRAE to turn experience into tools, turn work methods into systems, let AI take on more execution work, and leave our own time for judgment, creativity, and truly important things.

In the future, not everyone will need to found a company.

But more and more people will have the opportunity to gain some of the capabilities that only a team could provide in the past.

Everyone can try to become their own boss.

And TRAE may be the first "digital employee" you hire.

This article comes from the WeChat public account "Phoenix Weekly" (ID: phoenixweekly), author: Zhang Yu, editor: Yan Ruyi, published by 36Kr with authorization.

Source

36Kr · AI