The alarm clock rang at 6:15.
The first thing Han Luyi opening his eyes wasn't to check his phone, but to bring up Sight screen.
The experience bar remains unchanged: Lv.3, 1%.
Same as before I went to bed yesterday.
Han Luyi stared at the thin line where progress was almost imperceptible, his mind replaying past upgrade experiences.
He leveled up from Lv.1 to Lv.2 by reporting Flash. He first discovered that fixing real-world vulnerabilities could earn him experience points. Then he discovered that Flash was fabricating data and selling user privacy. After exposing this scam, he directly gained 200 experience points to level up.
The situation was a bit different from Lv.2 to Lv.3. At that time, he concluded that he could only gain experience by helping people for free, so he chose to answer difficult questions on technical forums using his online name. Later, he created BugKiller, which was more indirect and gave him even less experience, without any prompts. However, as more free users used it, he would gain experience, and eventually, he would automatically level up when he had enough experience.
I've been at Lv.3 for several months now, and my experience points haven't changed at all.
He thought he had reached the maximum level.
But last night the experience bar suddenly started moving.
Han Luyi carefully recalled what was different recently: he had used Sight intensively for three days, depleting his energy bar several times. The amount was indeed large. However, he had also accumulated a considerable amount of usage over the past few months by using Sight to scan code, documents, and business plans, yet his experience bar remained completely unchanged.
Is it because of this usage?
Or is it because training has started?
Han Luyi wanted to check the system records to see exactly when the experience points were added.
Sight: This function is not available.
Han Luyi then asked Sight, "Brother Sight, how do I advance to the next level?"
There was no response from the Sight.
This game is garbage. The game tutorial is terrible. The designer should be shot for five minutes.
After thinking for a long time without any clue, Han Luyi jumped out of bed.
Su Niannian specifically told him not to be late because she had a meeting with Neus at eight o'clock.
While washing up, a question suddenly occurred to him: video call, Ryan was in San Francisco, at least nine thousand kilometers away.
What can you see in the Sight?
Han Luyi arrived at the company a little after seven o'clock.
Zhang Biao went to park the car, Han Luyi bought an Americano and a latte downstairs.
Su Niannian arrived after waiting in the conference room for a while. She had put on exquisite makeup, but she still looked a little tired.
"Didn't sleep well?" Han Luyi said, handing the latte to Su Niannian.
"Thank you." Su Niannian took the coffee and placed it on the table, then covered her mouth with her other hand and yawned. "I went through Neus's information again last night, just cramming. You've been busy these past few days, so you probably haven't had time to look at it yet. Take a look first."
Seeing her yawn, Han Luyi couldn't help but yawn as well.
Su Niannian laughed, "This is contagious?"
Han Luyi took the tablet from Su Niannian. It had a Lark page open, containing information that Su Niannian had compiled from official websites and media interviews, with Su Niannian's annotations at the bottom.
"Neus AI, headquartered in Silicon Valley, was founded in early 2024. It received seed funding from Y Combinator and has raised funds annually since then. The most recent round was a $580 million investment led by Foundry Capital in July 2026, and it is now valued at $5 billion. The team has fewer than 100 members, mainly AI researchers and some senior programmers, most of whom come from well-known Silicon Valley companies. In addition, there are senior professionals from various fields responsible for alignment work."
(The Series B funding round only diluted the price by about 11.6%, indicating that investors have extremely high hopes for Neus. At least several companies must have bid for such a low dilution rate in the Series B round.)
"CEO Ryan Miller, 41 years old, has an impressive resume. He holds a PhD from the University of Toronto, where he studied under AI godfather Hinton. After graduating, he joined Google Research Institute and was part of the early days of OpenAI. He rose through the ranks to become a senior executive before leaving after the Altman ouster incident in November 2023 to found NeusAl."
(From researcher to executive to entrepreneur, it takes considerable skill and business acumen; be careful not to be blinded by academic prestige.)
Upon seeing this, Han Luyi couldn't help but click his tongue in amazement.
You wouldn't believe it until you see it; Chun Chun is practically an AI royalty!
"Neus AI's current flagship product is their proprietary Large Model, Neus Mercury, which has achieved top-tier performance in multiple tests, especially in code generation, where it has a unique advantage and enjoys a good reputation in the tech community. Their main revenue streams are API fees and subscription revenue from their web chatbots."
(API + subscription is the standard approach for AI companies, but the profit margin isn't high. Their main purpose in seeking cooperation with us is probably to expand their product line and complete their product ecosystem.)
Han Luyi nodded, agreeing with Su Niannian's assessment.
Source Code Technology's strongest asset right now is its accumulated product experience.
BugKiller's detection engine is a great complement to code generation, but generally speaking, the higher the accuracy of code generation, the lower the demand for BugKiller.
"We still have some time, how about we go for a run and try it out?" Su Niannian opened her laptop and asked Han Luyi.
The Neus Pyground page is now open.
"Have you already taken the test?"
"Last night," Su Niannian turned the screen around to show him, "You should take the test yourself too."
Han Luyi didn't rush to start; instead, he opened the Sight and performed a depth scan of the Neus model's code generation output.
Then he typed a line in the input box: Help me create a customer management system.
Press Enter.
Mercury responded quickly; the cursor blinked twice, and then the code started pouring out. First came a description of the project structure, followed by a series of code blocks: React frontend, Node. js backend, PostgreSQL table creation statements, with filenames clearly labeled and comments entirely in English—very professional.
Han Luyi page; the code for access control, data export, and email marketing modules was all there.
A chat window, over twenty huge blocks of code, and a complete Customer Relationship Management (CRM) system.
Looking at the Sight, there were only a few bugs, and they weren't in the core modules, which shows a high level of skill.
However, there are two problems: the entry barrier is too high, and there are too many things involved.
When a user mentions customer management, Mercury actually generates a complete CRM system, but the user still needs to manually deploy the front-end, back-end, and database to the Server. If the user is a salesperson at a small company, the software's two problems become two major challenges: they have the code but don't know how to run or deploy it; and many of the generated features are unusable.
This basically means it's unusable.
Han Luyi imagined the scenario: a salesperson holding this pile of code, Mercury telling him which website to go to, how to register, and which button to click to deploy... But would he really do it? Most likely he would just close it and download a ready-made spreadsheet template.
Han Luyi turned off his Sight and understood what was going on.
Mercury's code generation capabilities are solid, with well-structured and standardized code; its quality alone rivals that of Heavenly Craft. However, in terms of end-to-end user experience, they lack understanding of user needs, providing just enough functionality, and ensuring customers can use it immediately. The barrier to entry is too high, hindering its widespread adoption. In other words, they don't lack the Heavenly Craft, but they lack the Open Things.
The core of this advantage is what Open Things has been doing since its inception: enabling users to obtain the final product without needing to understand coding. He then thought of another aspect.
This only covers the code and tools. If a new model is developed, it will be able to solve intent understanding in all scenarios.
That's a market of a completely different scale.
He stopped thinking about it and turned to Su Niannian: "I see you have expectations about what he wants to talk about, and I agree."
"His main purpose in contacting us was to Open Things his product experience with Kaiwu. As for the specific cooperation methods, we'll hear what they have to say first. But we won't promise anything."
Su Niannian nodded. "Since it's a collaboration, we need to see what benefits they can offer first. No matter how good their model is, it won't mean anything to our current product line."
"Yes, just ask a lot of questions. I've heard that foreigners are all naive and rich."
Su Niannian laughed, "You've been watching too much Douyin."
At 7:58, Han Luyi clicked on the meeting link.