Chapter Six: Commitment — From Theory to Tool, How I Walked the Most Stubborn Path into the AI Era

Chapter Six: Commitment — From Theory to Tool, How I Walked the Most Stubborn Path into the AI Era In 2022, I called out to Yu Liang, wrote letters, made phone calls, and sent faxes. Then, it all sank like a stone into the ocean. But that outcry made one thing clear to me — I am not the only one who can identify problems, but too few people are willing to solve them. What I needed to do was not to keep calling out, but to build the tools that would solve the problems. So, I stopped talking and started building. I. Why College Students? In August 2020, I formally assembled a team. I recruited 15 people, all college students — seniors, third-year vocational students, interns about to graduate. At the time, the country happened to have policies encouraging the employment of college graduates, so I provided them with full social insurance and housing fund coverage. Why not hire experienced financial professionals? Traditional financial professionals simply could not understand what I was talking about. Their thinking was rigid. They lacked cross-industry thinking, management thinking, and systemic thinking. They could not even correctly map a company's business logic, nor accurately calculate the return on equity formula. But these college students, they were willing to learn. As long as you were willing to teach, they were willing to learn. Back then, I taught almost from morning till night — during the day, they wrote reports and searched for data; at night, I gave lectures, often lasting until one or two in the morning. They validated; I provided the ideas. That was my "Sunshine Team." II. The Limits of Excel By 2021, we had produced over 300 corporate diagnostic reports. But back then, producing a single report took at least half a day — manually downloading data, manually extracting accounts, manually copying and pasting into Excel, manually calculating revenue ratios, manually plotting trend lines. Excel can calculate anything, but it is far too limited. Employees could not log into a system; all data needed to be manually aggregated and uploaded, making it impossible to scale. Trend analysis, atomic task quantification — these logics could all run in Excel, but they could only reside on one person's computer, unable to be turned into a tool, unable to be used by more people. Later, we developed an automated tool and deployed it on Alibaba Cloud, but the operating costs were too high and simply unsustainable — the product had no promotion, no monetization, and the project ground to a halt. III. Committing All Resources The period from 2021 to 2023 was the most intensive phase of commitment in my life. Every month, salaries, rent, and server fees flowed out like a torrent. I committed all my resources, doing everything in my power to keep the team running. During those days, I tried every possible means to keep the project alive. The team shrank from 15 people to 5, and I could no longer afford the office rent. I had no choice; the project was halted. Then I went to Shanghai, seeking teams to collaborate with. But no one could understand this logic. I had the core parts of this system evaluated by Baidu; the other party said that developing just this one system would cost at least over 10 million yuan. I had no way to raise that much capital, so I could only temporarily pause and search for other avenues. IV. Stalled, But Not Giving Up From 2024 to 2025, I was in a state of stagnation. I was still persisting, still researching and developing, but unable to form a tool — it remained theory. In August 2025, I published a book — Financial Thinking Every Manager Must Learn. The book came out, and many spoke highly of it. But I knew that was not my endpoint. My endpoint was a tool, a tool that enterprises could use themselves, an implementable tool. A book can only convey thought; software can serve thousands, even millions, simultaneously. V. The Turning Point: AI Writing Code On December 16, 2025, a thought suddenly struck me: could AI be used to write code? That night, lying in bed, this idea suddenly popped into my head. I got up, opened my computer, and started experimenting. I used Doubao to develop my first mini-program — "Return on Equity Decomposition." It was my "debut work," taking about 20 days to go online. Looking back now, it was very rudimentary. The mini-program had black boxes, a fixed format, no room for innovation, a complex user interface, and a poor experience. But it went online, and it is still active today. That was my first step. Later, I learned about Python — an open language of building blocks that can be infinitely extended. Develop a function today, continue connecting the next one tomorrow, with no boundaries. I chose Python. From January to February 2026, I started developing with Python and built my first true system. Thanks to the rigorous logic accumulated from the previous 300 Excel reports, AI was able to convert the code so efficiently — my thinking was clear; AI merely turned it into a runnable program. At the time, I used local development, and then deployed it online. But the deployment process nearly broke me. VI. 432 Failures I chose a deployment platform called Zeabur. It was my first time deploying a system. I spent a cumulative 36 hours — over 10 hours a day, lasting several days. Every day I sat at my computer, started the deployment, waited minutes, failed, reviewed the logs, modified the code, redeployed, and failed again. One failure every 5 minutes, each failure accompanied by an email notification. 36 cumulative hours, once every 5 minutes — 432 failures. I scrolled through my inbox, densely packed with failure notifications, and the feeling was: you clearly know this thing can succeed, yet you are beaten back time and time again — 432 times. I used every AI tool available: Doubao, Qianwen, DeepSeek. They all advised me to give up: "Switch to another platform." But I stubbornly refused. I thought, if others can get their systems up, why can't I? If it doesn't work, I'll just modify the code. What's impossible about that? At the very last moment, the deployment succeeded. I will never forget the feeling at that moment. I stared at the success prompt on the screen, stunned, for a long time. Because of that experience, I can now deploy any system in seconds. Those failures transformed into the capability I possess today. VII. All the Systems I Developed From March 2026, I formally entered the full-scale development phase. To date, I have developed 10 systems, all deployed at aurum.zeabur.app. They can be divided into four categories: Category 1: Shengxin Investment Work Systems (3 systems) Aurum Personal Brand Showcase & Business Matching Platform, Shengxin Investment Official Website, NervOS Investment Manager Collaboration Management System. Category 2: Talent System (1 system) Woos Elite — a system aimed at evolving talent toward QS Top 50 universities. Category 3: U.S. IPO Cultivation Production Line Systems (5 systems) Wu ROE Financial Penetration System, Wu Enterprise Competitiveness Ranking System, Wu Financial Duel Diagnostic System, Wu Capital U.S. Stock Simulation Trading System, Wu U.S. Stock Ranking System. Category 4: Core System (1 system) Pureroot System · Bionic Human Value Engine. All those diagnostic systems preceding it were preparations for this single endeavor. And the Pureroot System strings all of this together into a complete production line — from enterprise diagnostics to system custody, from insurance-linked empowerment to capital matchmaking, ultimately leading to a U.S. IPO. These 10 systems were all written by me using AI. No outsourced team, no multi-million-dollar budget. Just me, and my AI employees. One might ask: are these systems theories that remain on paper, or products that have truly been implemented? My answer is: not only are they all online, but they are also running with real data. More importantly, they are not toys, but core assets with legal force. Currently, I have obtained 5 software copyrights, with another 3 patents in application. Moving from formal examination to substantive examination means the National Copyright Administration has already verified their absolute originality at the level of underlying code and logic. Without originality, one simply cannot pass this stage. Therefore, these 10 systems are not my "concepts," but industrial-grade products that I built with my own hands — already implemented, already in operation. VIII. Valuation How much are these 10 systems worth? I commissioned a professional appraiser to conduct a line-by-line evaluation. In 2024, Baidu assessed that the prototype partial functionality of the Pureroot System alone would cost over 10 million yuan. The complete 2026 version of the Pureroot System has a functional volume 3.3 times that of the earlier version. Below is the evaluation summary for the 10 systems: System Name Appraisal Conclusion (10k RMB) Reasonable Pricing (10k RMB) Aurum System V1.0 6–8 7 Shengxin Investment Official Website 3–5 4 NervOS Investment Manager Collaboration Management System V1.0 5–6 5.6 Woos Elite V1.0 13–15 14 Wu ROE Financial Penetration System V1.0 14–16 15 Wu Enterprise Competitiveness Ranking System V1.0 14–16 15 Wu Financial Duel Diagnostic System 3.0 75–90 80 Wu Capital U.S. Stock Simulation Trading System V1.0 8–9 8.5 Wu U.S. Stock Ranking System V2.0 14–16 15 Pureroot System · Bionic Human Value Engine V1.0 250–300 275 Total 402–481 439 The total appraisal value ranges from 4.02 to 4.81 million yuan, with a reasonable pricing of approximately 4.39 million yuan. Among them, the Pureroot System alone accounts for over 60% of the total value, making it the undisputed core asset. These appraisal reports are not fabricated estimates. Each one was broken down module by module, person-day by person-day, and expert cost by expert cost using the Function Point Analysis (FPA) method, conforming to the national standard GB/T 36964. The complete reports are appended at the end of this book. IX. The AI Era: Change Your Thinking I spent almost no money developing these 10 systems. I used AI for everything — DeepSeek for writing code, Qianwen as the architect, Doubao for graphic recognition and stack trace interpretation. They are my best employees: no salary needed, no rest required, available on demand. Now I have 5 software copyrights, with 3 patents under application. The Enterprise Competitiveness Ranking System I developed went online in just 13 hours. This would have been unimaginable three years ago — something costing 10 million yuan was completed for free using AI. I want to tell all entrepreneurs: in the AI era, everyone's thinking must change. All the old methods are different now. Society is operating at high efficiency; if you cannot keep up, you will be trapped. The AI era has shattered all rigid frameworks. It makes everything highly efficient. I walked the most stubborn path to arrive at the AI era. And the AI era allowed me, at the lowest cost, to turn the most complex system into the simplest tool. From 300 Excel reports to 10 systems; from committing all resources to free AI employees; from 432 failures to a 13-hour launch; from Baidu's 10 million yuan estimate to a summarized appraisal of 4.39 million yuan — I created a system cluster worth 4.39 million yuan at almost zero cost. I have walked this path. Now, I want to tell everyone still on the road: without money, you can still achieve great things. As long as your mind is not caged.