At 8:45 a. m., Jiang Yixin placed the jianbing (Chinese crepe) and soy milk she bought on the way to her workstation, then went to the break room to get a bag of potato chips, sat down in her cubicle, and happily began her "efficient early riser routine."
She ate her breakfast while checking her unread messages on Lark.
The entire screen is filled with it.
This wasn't a barrage of calls from some colleague who was always busy with online tasks, but rather Lark robot she had created herself.
The robot's name is "Eight A.M. Digest".
This is a small tool that Jiang Yixin created using Agent last week through "atmosphere programming".
Although she has now transitioned to product manager and no longer submits business code to the company's main repository, she is, after all, a top student who graduated with a degree in computer science and a product manager who transitioned from development, so her technical skills are still sharp.
After she and Qian Xiaole created the programmable Agent, she herself became a heavy user.
Back in college, Jiang Yixin would occasionally create a little productivity tool for herself.
When she was in college, she wrote a script for registering for online courses.
Back then, the school's academic affairs system was like a ticket hall during the Spring Festival travel rush on course registration days. The page would start lagging half an hour in advance, and at the very last second before class ended, the dormitory buildings were filled with the sound of refreshing the page. Popular general education courses were the hardest hit. Courses like film appreciation, introduction to psychology, and wine culture would show zero remaining spots as soon as they were released, before the page even loaded.
At first, Jiang Yixin diligently followed everyone else's lead, using three browsers simultaneously on her phone and computer, keeping her eyes glued to the countdown and her fingers hovering over F5. However, for two consecutive semesters, she only managed to snag a bunch of classes on Friday evenings and Monday mornings at 8 AM.
Before the third time she tried to register for a class, she finally couldn't take it anymore.
Before I started learning computer science, I diligently registered for classes; now that I'm learning computer science, I'm still diligently registering for classes. Did I learn computer science for nothing?
It just so happened that they were learning about cybersecurity in their major courses, Jiang Yixin put all the new knowledge she learned in class to use.
She first spent a weekend studying the school's academic affairs system, capturing network packets to view requests, and then analyzing login status and CAPTCHA refresh logic. She then painstakingly extracted the course selection interface, course IDs, submission parameters, and failure return codes.
The most troublesome thing was that the academic affairs system was poorly written. Sometimes the same operation would return JSON, and sometimes it would return a piece of HTML. The error message was also hidden in the page script. She worked until 3 a. m. for several days in a row to get the whole process running.
On the day of course registration, she didn't frantically refresh the webpage. Instead, she filled in the course ID into the configuration file beforehand. When the time came, the script automatically logged in, checked the remaining course slots, and submitted the course selection. Three seconds later, a message popped up on the screen. "Course selection successful."
Later, the script was passed on to classmates—first to Big Liu, then to his other roommates, and eventually to the entire computer science department.
Later, it became clear that students from other majors were complaining that the film appreciation class was being taught entirely by computer science students.
The matter eventually escalated to the Academic Affairs Office.
The professor who taught them cybersecurity was called in by the academic affairs office to help, and spent two days rewriting an anti-scripting mechanism for the academic affairs system.
Then Jiang Yixin was called in for a talk. She was trembling with fear, thinking she was going to be punished, but to her surprise, after the professor saw her, he immediately put her code to the test and asked her how her script should be protected.
In the end, not only was she not punished, but she also received full marks for her class participation.
This incident made Jiang Yixin famous overnight, and she became a well-known academic star in the computer science department.
She's still quite proud of it when she thinks about it now.
But even writing a small tool used to be physically demanding.
Only a small portion of the projects can be written down; most projects fail halfway through their development.
Especially with tools for personal use, the need arises quickly and disappears just as fast. Often, she'll enthusiastically create a new folder, writing a README that's more complete than the code itself, only to start questioning her life choices by the time she gets to the third exception handler.
The tool wasn't even finished before the matter was resolved.
The folder is now filled with a bunch of half-finished projects with exciting names but pitifully low completion rates.
But with Agent, everything changes!
Once she breaks down the requirements clearly, the remaining code can be directly generated by the Agent.
Finally, she looked at it again herself, made changes to a few things that didn't feel right, ran two tests, and finished building it in a little over an hour.
This feeling is kind of intoxicating!
She can easily come up with a small idea to improve efficiency and integrate it into her familiar workflow. She can also take over any permission issues that the Agent cannot resolve on its own.
This feels different from using Open Things. Open Things focuses on generating apps for users without a technical background, with a built-in interactive interface, ultimately presenting the interaction between the app and the end user.
However, Agent are designed specifically for programmers and are more customizable. Jiang Yixin can customize her interactions with external information.
For example, this "Eight A.M. Digest" robot.
Every morning, it automatically scrapes several publicly available information sources from External Network: TechCrunch, The Verge, Product Hunt, and other Tech Media, several AI-driven social media accounts, as well as some keyword pages that she manually adds.
If the title or body text contains keywords such as Nexus, Kaiwu, or AI dev tool, the content will be summarized using AI, and the link will be pushed to you via Lark robot every morning at 8:00 AM.
For a product manager, this is sufficient.
The backend data can tell Jiang Yixin what her users are doing.
This little gadget, however, alerted her to what the people outside were doing.
After her promotion, Jiang Yixin was responsible for overseeing all product lines at the Source Code company, managing three product managers. She had to follow up on product decisions and also handle some of the collaboration with Nexus. Therefore, she couldn't just focus on the backend data; she also had to keep an eye on overseas public opinion and industry signals.
Typically, the "Eight A.M. Digest" pushes out three to five news items per day.
But nine items were recommended today.
The first few points are all fairly standard.
One story is that a user on X shared their experience using Kaiwu, saying they created a family budget management tool with Kaiwu. Many people are forwarding it and asking for links, so the original poster took the opportunity to release a paid version.
One suggestion is that there's a trending topic on Product Hunt, and Kaiwu should add more templates.
Another point is that TechCrunch's weekly AI product report mentioned Kaiwu, saying it's a rapidly growing AI app building tool.
All the news was good. Jiang Yixin quickly glanced at it and prepared to continue reading.
Then her gaze settled on the seventh item.
This is a summary of a paywall news item from a Tech Media.
The title is very long.
Nexus begins testing an AI development tool for non-technical users.
Jiang Yixin clicked on the summary.
The gist is that Nexus is internally testing an AI development tool aimed at non-technical users, including Small and Medium-Sized Enterprises owners, independent developers, and content creators.
On its own, this is nothing special.
For a unicorn company of Nexus's size, it's normal to internally test an AI tool.
Jiang Yixin continued reading.
The eighth item is from LinkedIn (the international version of LinkedIn).
It's a personal update, posted by David Chen.
He comes from a major Silicon Valley company and has been working in the AI product department at Nexus for over two years.
Three days ago, he posted a very short status update.
[Exciting new project. More soon.]
Someone commented below asking, "Is this related to development tools?"
David Chen simply replied with a smiley face.
That's nothing.
It's all too common for employees of big companies to post these ambiguous updates on LinkedIn, just to pique your interest, and the final product is often not very good.
Jiang Yixin continued reading the ninth item.
This is an AI industry report published by a self-media blogger, summarizing the recent recruitment activities of AI companies over the past week.
One of them was a position at Nexus.
[PM, AI Creation Tools]
The job description includes the following sentence: Familiar with the usage habits of non-technical users in the Asia-Pacific region, and has relevant management experience with AI tools.
Jiang Yixin stared at that sentence, along with the Tech Media news article and the Nexus employee's LinkedIn post.
Each of the three lines individually seems insignificant, but when put together, they convey only one meaning—Nexus wants to create its own Kaiwu.
Moreover, they have also set their sights on the Asian market.
Jiang Yixin narrowed her eyes, her gaze sharpening.
She knew her contract with Nexus expired in April, and she also noticed that the people on Nexus's side were becoming increasingly ambivalent towards Kaiwu.
While continuing to obtain real-world user data through Kaiwu, they increasingly emphasized "phased cooperation," "data sharing boundaries," and "keeping the future product direction open" in their communications, mentioning less and less long-term planning.
Previously, these things only made Jiang Yixin feel uncomfortable. Now she knows where that discomfort comes from.
Jiang Yixin was a little flustered.
After calming herself down, she first sent Su Niannian a message via Lark:
"President Su, are you free right now? There's something I'd like to discuss with you."
A few minutes later, Su Niannian replied:
"Come to my office to talk."
Jiang Yixin unplugged the data cable from her laptop, picked up the laptop, and headed straight for Su Niannian's office.
"What is it?" Su Niannian looked up from behind her desk.
Jiang Yixin placed the screen directly in front of her.
"Please take a look first."
She felt it was difficult to explain in a short time, so she suggested letting Su Niannian look at the original information herself and see what her opinion was.
Su Niannian read through the messages one by one. After finishing, Su Niannian remained silent for a while without speaking.
It's as if all the clues have been gathered and presented to the detective.
Su Niannian didn't give an answer, but instead asked Jiang Yixin:
"Xiao Jiang, what do you think?"
Jiang Yixin stated her own judgment. "They want to create their own Kaiwu."
She paused for a moment, then added.
"They want to kick us out and take over our market."
As she said this Jiang Yixin couldn't hold back her grievances any longer. She had recently put in a lot of effort to localize the product for the United States market, researching many tools and spending her days on Reddit (the overseas version of Baidu Tieba) reading user feedback.
The user growth curve is currently rising, and the conversion rate and repurchase rate are both very good. I never expected to be betrayed by a partner at this crucial moment.
Su Niannian glanced at Jiang Yixin's expression but didn't rush to offer any words of comfort.
She scanned the three messages on the screen again, closed the laptop, and pushed it towards Jiang Yixin.
Jiang Yixin looked at Su Niannian, not understanding what she meant.
Su Niannian thought for a moment and said, "Boss Han is still upstairs with the model team. I have a meeting with him this afternoon. Why don't you come over with me?"
It sounded like they were going to discuss a strategy, but Jiang Yixin vaguely felt that Su Niannian wasn't as surprised as she had imagined.
Jiang Yixin nodded. "Okay."
Just before leaving, Su Niannian called Jiang Yixin back, smiled at her, and said, "Don't worry, everything will be alright."
Jiang Yixin carried her laptop out of the office.
He was still thinking about the astonishing conclusion he had just deduced.
She initially thought she had caught a fox's tail.
This is Nexus AI's true intention hidden behind the partnership.
Logically, now that the detective has found a clue, the case should move forward.
What surprised her was that Su Niannian didn't panic at all after seeing the fox tail.
Jiang Yixin: She must know something I don't.
This feeling is not very comfortable.
For the first time, Jiang Yixin found the riddle-solving CEO President Su a bit annoying.