Candid Talk on English, Life Planning and Personal Finance in Tech | Original
Two participants: Zhankang and lzwjava.
Han Zhankang, entrepreneur, 10 more years in backend engineering and archticture. WeChat@laiceshide
The following conversation was transcribed using Whisper on an RTX 4070, then refined and organized with the help of AI tools. The original conversation was in Chinese.
Note: Because both transcription and refinement were AI-assisted, some details may be inaccurate or paraphrased. Please verify any important information independently before relying on it.
AI Translation tips: Han Zhankang，韩占康。
Candid Talk on English, Life Planning and Personal Finance in Tech
This conversation centers on two senior tech practitioners discussing their career development, technical exploration, English learning, financial situation, and life planning. Both candidly shared their journeys from high-paying positions to entrepreneurship, from deep technical specialization to going overseas. They explored in depth the technology stack choices in the AI era, the critical role of English ability in the workplace, the cutting-edge trend of combining hardware and software, and how to maintain mental balance and continuous growth during an economic downturn. The conversation also touched on specific interview experiences, open-source tool usage, command-line workflow construction, personal financial management, and family relationships.
Career Development and Salary Discussion
From High Salary to Entrepreneurship
One participant shared their salary trajectory: at their last job, the annual salary was just over 1 million RMB, especially during 2023-2024. But after starting a business, income became unstable, and now when job-hunting, salary expectations are hard to bring back to the previous level. He worked for about 4-5 years, mainly with a backend technology background, moving from expert to architect roles. The other participant mentioned that on the market, many companies can afford to pay “skilled people” 1 to 2 million RMB annually — the key is one’s own market value positioning.
Salary and Real-Life Costs
The discussion dove into the actual take-home pay of high salaries. After fully compliant tax deductions, a 1 million RMB annual salary takes home about 500,000-600,000, averaging just over 40,000 per month. This means that while it looks glamorous on the surface, after deducting mortgage and daily expenses, one can only save 100,000-200,000 a year and buy some stocks — actual disposable funds are not much. One participant commented: “It feels glamorous on the surface, looks like a million, but a large portion is actually taken away by taxes.”
Regional and Industry Disparities
The two discussed the salary polarization in China’s IT market. In the US, an average engineer earns around 100,000 USD per year, but the salary gap in China is enormous — poor engineers may earn only 2,000-5,000 RMB, while top talent can earn millions or more. In Guangzhou, there are at least a thousand, or even a hundred, companies offering high salaries. But in recent years, except for AI-related companies in Beijing (like Kimi and robotics companies) that have raised funding, other companies are generally doing poorly. Giants like ByteDance, Douyin, Meituan, Didi, and JD are all contracting or laying off staff.
Technology Stack and Work Content
Actual Work in the AI Field
One participant mainly works in AI on training and inference, while also doing AMD-related open-source projects. He fine-tunes models less and focuses more on training and inference. The other participant works on AI application development in fintech, focusing on upper-layer development using commercial large model APIs, without needing to go deep into underlying training and inference. His annual token consumption mainly comes from this application development work.
Backend Architecture Reflection
Regarding microservices architecture, one participant offered a core view: microservice partitioning is closely tied to the business and requires domain identification, microservice implementation, capacity planning, and performance optimization. But in his view, these are all “implementation work” and “pseudo-construction” tasks, no longer the main focus today. Any architecture (microservices, monolith, etc.) is fine — the key is service decoupling. He elaborated on the trade-off principle: a monolith works at small scale, but as scale grows it should be gradually split into microservices. Whether to split into 3-5 or 50 microservices depends on the actual business situation.
Experience with Code Generation Tools
One participant mainly uses the Codex model, finding it task-friendly, not prone to randomly banning accounts, and with acceptable usage quotas. He buys two or three accounts to rotate, because the $20 account has usage limits. These accounts are mainly bought from Xianyu (many people use this channel), while the other participant has bought them from the official website (requiring a US credit card or Hong Kong payment method). Both agree that Xianyu is cheaper.
Evaluation of AI Code Generation
Both agreed that current AI code generation is already “pretty good.” The main issue is that the content is not accurate enough — sometimes it “goes off track” and requires good prompts to guide it. Accuracy still needs improvement, and prompt quality directly affects output. One participant commented: AI has already exceeded his own coding level, so he no longer needs to write much code himself.
English Ability and Career Opportunities
The Critical Role of English in Interviews
One participant shared his experience: he once got an interview opportunity at Binance through an internal referral, but ultimately failed because of English. He commented that Binance’s Java interview was not difficult, but English requirements were high. The other participant works as an outsourced employee at an international bank (30+ years of work), with fluent spoken English, able to chat daily. He previously worked as a contrator at DBS Bank, took the IELTS, used English at work, and even communicated with Indian colleagues.
Self-Assessment of English Level
In the latter half of the conversation, one participant demonstrated his spoken English by impromptu answering the question “What do you do today?”, describing how he discussed architecture with colleagues that day and planned the migration of local Redis to the cloud. Although he rated his own spoken English as “very simple, not standard enough,” the other participant thought it was already quite good.
English Learning Strategy
One participant shared his English learning experience: he scored 8.5 on IELTS Reading, with an overall score of 6.5. He believes he can walk into the exam room and get this score anytime without preparation. His experience is based on actual usage scenarios, including using English at a foreign company and reading large amounts of English technical articles in his spare time. He wrote an English learning guide of 5,000-10,000 words, bought supplementary courses for about a dozen RMB on Pinduoduo, and improved through reading many English blogs.
Englishization of Work and Lifestyle
One participant described how he “completely Englishized” himself: not reading any Chinese content, only reading English technical documents and blogs, taking junior college exams in English (all subjects). His IELTS Reading score of 8.5 lets him scan English articles as fast as Chinese. Now, in Zoom meetings, he can go months without subtitles when chatting with Indian colleagues. But for meetings with other teams or his boss, he turns on real-time translation (Zoom’s real-time translation feature) as a safety net.
Layered English Learning Advice
One participant proposed a layered strategy: in familiar team meetings, deliberately turn off subtitles to practice listening and allow yourself to ask the other person to repeat once or twice; in meetings with unfamiliar people, turn on subtitles to gradually adjust your comfort zone. He emphasized: “If you make others repeat three or five times, no one has the patience.” He also suggested fully Englishizing your life — reading, watching videos, thinking, and communicating all in English. The first month or two will be painful, but you’ll quickly get used to it.
Financial Situation and Family Management
Asset Shrinkage and Mortgage Pressure
One participant described his financial situation in detail: total family assets dropped from a 2021 peak of 3.5 million to less than 2 million now (possibly lower in reality). The main reason is a housing loss — he bought a house in Beijing for 3.1 million, which may now be worth only about 1.8 million. He took out a loan of over 2 million, which has been paid down to 600,000 remaining, but the house can’t be sold. Monthly mortgage pressure is heavy — almost everything beyond daily necessities goes to the mortgage.
Family Asset Structure
Current family annual income totals about 500,000-600,000 RMB (Guangzhou). If he can get a higher outsourcing salary, it may reach 900,000. He is 33, his wife 31. His parents own a home, but total family assets have been halved, mainly due to a housing loss of about 1-2 million. One participant pointed out that prices are falling everywhere in China — almost no one is making money, and the more you buy the more you lose.
Family Living Arrangements
One participant has a somewhat unusual family living arrangement: on weekdays he lives separately from his wife and children. The kids attend kindergarten near their grandparents (Huangpu District), while he lives in Zengcheng District. He goes to the office three days a week and occasionally passes by his parents’ house to see the kids for an hour or half. On weekends, his wife brings the older daughter over while the younger daughter (2 years old) stays with the grandparents. This arrangement provides him with some independent space and reduces conflict with his father, who used to often say he was “not doing proper work” because of his hobbies with circuit boards and his education level (junior college). He spent 2 million RMB (including 1 million lost on the house) to live there, and now he and his father haven’t argued in a long time.
Entrepreneurship and Side Projects
Live Streaming and Brand Building
One participant shared his entrepreneurial experience: he once did live streaming, accumulating connections through gifts and meeting many iOS gurus, building his brand well. But after switching from iOS development to full-stack backend, he became disconnected from the old circle. He commented that he had “lost touch with the people” — as an iOS engineer, his traffic and users were mainly in the iOS circle, but after switching to backend he lost these resources.
The Difficulty of Outsourcing
He mentioned he had taken on several outsourcing projects before, which were “exhausting.” Of 50 outsourcing projects, about 5-6 were “really bad” — months of work for just 10,000-30,000 RMB. The other participant remarked that the high threshold for opening a bank account (100,000-200,000) makes sense — it indirectly reflects how hard it is to make money domestically.
Lessons on Funding and Management
One participant recalled that if someone had invested 1-2 million in him, he might have grown quickly. Later someone did invest 500,000, and he immediately hired people, but discovered he didn’t understand management. He reflected that if someone invested 500,000 now, he wouldn’t move it (would put it in Bitcoin or something), then publicly claim to have money, and activate resources through various free relationships (sending red packets in group chats, sharing knowledge). He now often sends red packets and shares knowledge. No one complains about his red packets being too small anymore (in the past someone complained that 1 RMB was too small), and old friends know he won’t cheat them and has always worked hard.
The Gap Between Product and Marketing
One participant acknowledged his dilemma: he doesn’t lack product or market — he mainly lacks marketing. The product is already there, but customer acquisition requires constant traffic investment, and he’s not a professional in that area. It’s hard for him to change and put himself out there for marketing because he’s used to staying low-key. He feels comfortable in group chats but insecure in Moments — worried about telling others he’s at an outsourcing company with a salary under 300,000, in an awkward position (he had 380,000 at 27, and now things are getting worse).
Reflections on Industry and Era
Era Dividends and Opportunity Gaps
The two had a deep discussion about the era. Post-80s entrepreneurs (like Wang Xing, Zhang Yiming, etc.) were born earlier and entered society around 2003-2005. By the time the mobile internet boom hit in 2013, they were already mature. Combined with the capital boom, any business could easily acquire users. But post-90s (their generation), wanting to start businesses around 2013-2016, found the slots already taken — they were fresh out of school with limited understanding of the world. Now, only AI large model entrepreneurs with Tsinghua or Peking University backgrounds can succeed, but the bar is extremely high — founders basically need a PhD or higher.
Era Dividends and Personal Effort
One participant believes that people who master core business knowledge (like management, human nature) and have technical backgrounds (like Wang Xing, Zhang Yiming) can earn tens of billions — it’s not just personal ability but also the era’s problems and dividends. The other added that every era has its own story, and the current AI wave is an opportunity for post-90s and post-00s entrepreneurs.
Pessimism and Economic Conditions
Both felt an unprecedented downward trend in the job market. Ten years ago, you could just interview casually and easily get a 200,000-300,000 job. When he first graduated, his salary was 2,500 (Xi’an), and around 10,000 when he arrived in Beijing. Now, job hunting requires using all channels (Boss Zhipin, friend referrals) simultaneously, and even so it’s hard. One participant mentioned that a friend communicated with over 7,000 people on Boss Zhipin, with about 2,000-3,000 reaching out, but most large companies are not in good shape.
Discussion on Future Direction
Both recognized the transformation brought by AI. Work that used to take two months can now be done in a day. The current tech stack (large model training and inference) is more valuable for accumulating technical depth, while technologies like Harness AI or Agents abroad are already mastered by most engineers.
Hardware and Software-Hardware Integration Outlook
Layout in Chip and Robotics Fields
One participant shared his thinking on positioning in chips and robotics. He uses open-source tools (like Verilog — a hardware description language) to learn, because open-source lowers the entry barrier. Existing AI Agents can assist software development through text-based code, and for hardware, just connect devices like Arduino to a computer and AI can help write control code.
A New Paradigm for Hardware Development
He proposed a new approach: take phones and computers as the first layer of hardware, connect all electronic components to the computer somehow, then use AI Agents to quickly write code to control them. Connect every electronic part of a robotic arm, the main unit of a DJI drone, and the various parts of a chip to the computer, then assemble them. He especially emphasized “everything should be considered from the AI Agent’s perspective” because AI’s reasoning capability is so powerful.
Background in Software-Hardware Integration
One participant believes he has an advantage in software-hardware integration: strong hands-on ability (making porridge, building furniture, assembling computers), foundational circuit knowledge from junior college exams (capacitors, inductors, transistors), and familiarity with hardware devices like OpenWrt routers. He believes that if he can bridge software and hardware, he could become a “mini Zhihui Jun” (a well-known engineer skilled in both software and hardware) in the AI era.
Judgment on Future Tech Trends
The conversation pointed out that bridging software and hardware is the mainstream direction. Zhihui Jun was already ahead of his time three to five years ago — he understands everything (AI, software, hardware, robotic arms, self-driving bicycles). Now with AI, the bar for writing code is greatly lowered, and human creative value has become more important. For someone with software and hardware foundations, becoming the next Zhihui Jun or a mini version is not difficult.
Interview Experience and Job Hunting Strategy
Binance Interview Recap
One participant shared his Binance interview experience: he got the opportunity through a friend’s internal referral, but only one referral was allowed. After that he had to apply through the official website. He commented that Binance’s Java interview was not difficult — mainly the English requirement was high. The interviewer asked about IOC (Inversion of Control) and AOP (Aspect Oriented Programming) and other Java techs, which he could explain fluently in English (e.g., describing IOC as “you give the code to the library, the library calls your code”), but he ultimately failed because his English and reactions weren’t fast enough.
Lessons from the Standard Chartered Interview
The other participant shared his interview experience for a Standard Chartered Bank outsourcing position (annual salary over 400,000): he failed in the last round (the third of four), because the interviewer asked about RAG (probably some technical term) and Java fundamentals. He was a bit confused at the time — although he had actually done it (spent a day or two on RAG), he hadn’t gone deep, so he directly said he hadn’t done it. He demonstrated more in-depth content like GPU training and inference, but the interviewer only knew about Agents and wasn’t at his level on AI.
Reflections on Interview Skills
One participant advised: when job hunting, you need to “talk it up and stretch it” — never be modest. The other commented that the first participant’s AI-related work (inference tuning, etc.) was actually more than his own, but because interviewers weren’t familiar with those areas, he couldn’t get recognition. Both agreed that the AI field is developing fast — there’s still time to learn after getting in.
Personal Life and Mental Health
Depression and Mindset Adjustment
One participant candidly talked about his depression. His current mindset is “even if I die tomorrow, I’ve done what I needed to do” — he’s experienced marriage, having kids, and glorious moments in youth. Now, though things are bleaker, he’s leveraging AI to do something again. He explained that one reason for his depression was spending a lot of time thinking about the world — only after going through many experiences does one become more clear-headed.
Family Relationships and Independent Space
About family relationships, he described his living arrangement in detail: on weekdays he lives alone in Zengcheng District, on weekends he reunites with his wife and two daughters. This “solitude” brings independent space and avoids conflict with his father (who often criticized him over his education level). He considers solitude a good thing — family relationships are now more harmonious and they no longer argue.
Attention to Physical Health
One participant mentioned he weighs 96 kg and is over-nourished, so he often drinks porridge. The other suggested medications like Semaglutide, saying he’s also planning to buy it for his father — just hasn’t gotten to it because of recent interviews and work.
Future Planning and Action Intentions
Three-to-Five-Year Personal Development Plan
One participant laid out an ambitious plan: first get a regular position at an international bank, while taking the self-study undergraduate exam (he’s already passed the computer subjects but is a few points short on general courses like College Chinese and Marxism). Then move to Hong Kong or Singapore. By 2027-2028 (around age 36-38), present himself fluently in English on the global geek stage, do a creative project like Zhihui Jun’s, aiming for 100,000 followers. Then gradually expand to 500,000 or even 1 million.
Tech Investment Plans
He plans to replace all three of his phones with iPhone 16 Pro (second-hand) in the coming months, expanding to four phones. He already has five or six computers at home, including a local GPU environment. He plans to invest 50,000-60,000 RMB in stronger GPUs over the next year, gradually multiplying the number of local app installations tenfold (from 200-300 to 2,000-3,000). In terms of token consumption, he currently uses about 500 million per week, 2 billion per month, planning to increase to 5-10 billion per month in two months, and to compete with peers after sustaining this for three to five years.
Refined Operations Plans
He’s already building a super command-line toolkit (including networking, translation, blog publishing, screenshot notes, meeting transcription — around 100+ commands), but only he uses it. Next he plans to redevelop marketing command-line tools to automatically publish content to multiple social platforms (Xiaohongshu, Twitter, etc.). He’s registered about 30 social accounts and continues to expand. His blog currently has 50,000-70,000 monthly visits, about 500,000 yearly, and he plans to monetize through Google Ads.
Long-Term Layout for Software-Hardware Integration
He plans to start later this year or next, leveraging his existing hardware foundation (circuit knowledge, device assembly experience) and supported open-source chip and robotics projects (Verilog, Arduino, etc.) to bridge software and hardware skills. Ultimately, by 2027-2028, present comprehensive capabilities fluently in English on the global stage.
Conversation Wrap-Up and Emotional Exchange
At the end of the conversation, both expressed deep friendship. One participant said that when he was depressed, the other was the best among the dozen or twenty friends who genuinely cared. The other responded that he is also an entrepreneur, sees many excellent qualities in the first participant, and that life is full of experiences — even adversity is an experience. They encouraged each other, believing that as long as you find the right direction and persist, you can keep improving. Finally, one participant said he was going to drink porridge, while the other was going to do GPU training. They agreed to keep in touch.