Since graduating in 2011, I’ve lived in Guangzhou, Hangzhou, and Shanghai, but somehow ended up right back in Guangzhou. Now, as I near 35, something clicked (or maybe snapped). No matter where I live, what company I work for, or how much I earn, I keep spending over 12 hours a day chasing someone else’s dream. The rest of my time is divided between family, friends, health, entertainment, and — most importantly, yet often neglected — my own dreams.

At this age, in this economic climate, talking about “dreams” feels almost embarrassing. Without financial security, dreams can feel like nothing more than wishful thinking. You can talk and daydream, but reality pulls you back down.

Still, after long work hours led to sudden hearing loss, my desire for freedom broke through. With what energy I have left, I’m shaking things up — switching careers to become an independent developer. At the very least, I’m starting with a side project.

When I told a friend about my plan, he joked that I was entering the field at the worst time. While people talk about “35-year-old developers getting laid off,” here I am diving into coding. Online, some exaggerate, comparing learning to code in the age of AI to “catching the last train” or “joining the KMT in 1949.”

Add to that the tech bloggers forced into “layoff lanes” because of the economy, all working on the same “indie dev trio”: bookkeeping, to-do lists, and note-taking apps — competition is fierce. Even with AI’s help and a small stash of “F-You Money” to keep me afloat, leaving a steady income to become a near-full-stack indie developer feels irrational. Especially when 9 out of 10 indie devs are losing money. This isn’t a typical decision for someone nearing 35.

I had plenty of doubts. Within the first month, I regretted my decision more than once. Sleepless nights, stress-induced headaches, acne breakouts — all part of the package. I coped by binge-eating McDonald’s and playing Steam games. After 15 years of always having a paycheck, the sudden absence of one made me miss the certainty of my old life.

Working for someone else is exhausting, but it’s predictable. There’s a guaranteed return, even if it doesn’t always feel worth it. That certainty reduces stress and anxiety, helps you sleep better, and gives you the energy to show up and complain the next day.

But when that certainty is gone, the mental struggle begins. Whether you escape into video games, or dive into activities like weightlifting, drinking, running, or splurging on luxury experiences, you eventually come back to the same question: What kind of life do you want? What kind of person do you want to be?

For me, the shift from product manager to independent developer isn’t just a career change — it’s a mindset shift. Now, every decision, action, and result is my responsibility. This is the “one-person company” I’ve always imagined. Like any company, its limits are defined by the founder’s vision and mental boundaries. To succeed on this path, personal transformation is essential. Past achievements only prove I was a good cog in a machine. But once you’re on your own, the struggle is inevitable.

Still, I’m fortunate. Today’s endless learning resources — videos, documentation, forums, and of course, AI — make it possible to learn almost anything. The challenge isn’t in learning; it’s in how efficiently you learn. And there’s always a gap between “learning” and “mastering” that only hands-on experience can fill. No matter how many hot topics you follow, without practice, you’re just chewing on leftovers.

For me, the toughest part of indie development wasn’t the technical side. (As an example, I built my first product in three weeks, mostly with AI’s help — more on that later.)

The So-called Choice

The biggest challenge was choosing a direction. Most startups face the same dilemma. Is choosing a “track” really a choice? My first employer started with monitor PCBs because the team already had experience there. From that, they expanded into education and conference displays — just monitors in different settings. My second employer began with email, shifted to web portals, then games, and eventually e-commerce after going public on NASDAQ. His path mirrored the broader internet wave. My latest employer went from traditional car manufacturing to new energy vehicles. Was it really changing industries? Not quite.

It might look like a “choice,” but in reality, they were feeling their way forward. No startup works like Cities: Skylines in infinite money mode, laying down perfect road networks to avoid future problems. For me, all I can do now is find a roughly correct direction within my current resources and navigate step by step. I’m running a “one-person company,” and my logic has shifted.

I chose “HR” + “AI” + “to C.”

HR + AI is nothing new in the B2B world, and it didn’t yield much success before. But with the GPT wave, the concept got a fresh boost. Even so, after a year or two, practical applications remain limited to a few areas:

• Recruitment: JD generation, resume scoring, interview assistance
• Employee experience: smart customer service
• Talent development: talent profiling
• Efficiency tools: report generation and organization

From a business perspective, there are universal methods that product managers can use to build structures, no matter where they go. But once you move past the initial phase, you face diminishing returns. These applications mostly improve worker efficiency and user experience — they don’t bring direct revenue to companies. As for cutting labor costs? We’re not quite there yet. I haven’t seen an HR AI product that fully replaces a real employee. Even the most advanced, like Max in Silicon Valley, only assist with resume screening and interview scheduling. At best, companies might cut back on interns or junior staff, but the savings are minimal due to additional service fees.

Without experience handling basic tasks, how can a junior employee grow into a mid-level one? You can’t expect someone to be a good interviewer without reviewing at least 1,000 resumes.

For product managers in enterprise solutions, as they progress, their work becomes more about playing a numbers game — proving their value by “decorating a pile of crap.” But this can also be their moat, as long as they hold on. Just like legacy code, no one dares to touch it.

But this isn’t why I chose “C over B.” In my view, independent developers aren’t well-suited for B2B applications. First, I’ve left the collaborative environment where B2B product design thrives, so it’s easy to lose direction. B2B is about “process fit,” not “feature fit.” Even with great features, if they don’t integrate into workflows, their value is limited. Plus, B2B project cycles are long and operations are complex — ill-suited for solo developers.

So, I’m not “choosing” to avoid B2B — I just can’t do it alone. Most internal B2B processes are poorly standardized, while AI demands a high level of standardization. Enterprises have little tolerance for AI’s output errors, and pushing AI in this field requires more manpower to clean data and verify results.

Lack of trust between enterprises, data compliance issues, and geopolitical influences add further restrictions to non-edge AI models. Even if Model B is better than Model A, enterprises often spend more manpower to adapt the inferior option for compliance reasons. Internal competition within large companies exacerbates this. These are just a few unavoidable challenges of AI in the B2B sector. (Of course, some companies build AI SaaS platforms like Dify, but that’s a topic for another article.)

I’m not equipped to solve these challenges, nor am I qualified to dive deep into them. So I’m focusing on what I can manage: AI applications for consumers.

Consumer-facing AI applications come with their own set of challenges. While they can be developed quickly, they die just as fast. Most ideas seem like wishful thinking from product managers or business owners. Even the occasional breakout success lacks sustainability — eventually, costs, compliance, or the evolution of the underlying AI models swallow them up. Beyond FOMO and traffic-driven initiatives, the internet is flooded with “AI mentors” or “startup gurus” who haven’t succeeded themselves but are happy to tell you how to do it. (Honestly, I’m no better — I’m just venting.)

I’m not listing these challenges to give myself an excuse — well, maybe a little. It’s easier to save face if things don’t work out. But when you look at these difficulties from another angle, the real question isn’t, “Should I take this path?” but, “How should I walk this path?” The worst-case scenario is ending up with nothing, but I’ve set a stop-loss line. The important thing is to start.

Once I mentally framed it as, “Just give it a shot — if it fails, move on,” I was ready. Curiosity for new things, excitement for technology, and the relief of not having to justify product value in a bureaucratic company fueled me. When I started writing code, I couldn’t stop.

The Joy of Coding

Writing code gives me a satisfaction like nothing else — even more than scrolling through TikTok. Watching TikTok is a battle of self-control: “Just one more… okay, maybe three more… wait, how did it get to 1 AM?” The endless scrolling feeds the next dopamine hit. But coding? That requires thought. Even if AI writes most of it and I’m just playing the role of Ctrl+C or Tab Engineer, I still need to know where to paste without causing errors. And when errors pop up, I have to debug. All the pressure builds up before hitting “run,” and even more so when an error appears. But the rush when it finally works? There’s nothing like it — so addictive, so satisfying.

In product management, creating a polished prototype gave me a similar feeling, but even the best high-fidelity mockup wasn’t what the user would get in the end. Coding is different. Every line represents the final product — nothing is more “what you see is what you get” than this. It satisfies my control freak tendencies completely.

Of course, I’ve also realized that some of the clever designs from a product perspective don’t align with coding patterns and can be a nightmare to maintain. Now, I understand why developers say, “I’m just changing a field,” and then schedule a whole day for it. (We can discuss more on that in future blogs.)

Fueled by that sense of immediate satisfaction, I built my first indie product.

It’s a career guidance AI assistant for individuals. The services it offers include:

• Resume polishing: rewriting existing resume content based on target job responsibilities and requirements (already live)
• Resume builder, interview preparation, mock interviews (still in development)
• A few small tools to assist with resume writing
• The goal is to provide users with tailored guidance and advice throughout their careers.

With AI’s help (it wrote about 70% of the code, I wrote 10–20%, and the rest came from existing frameworks), I built the basic framework, the resume polishing module, and some small tools — all within three weeks. The other modules are still in progress. If you want to try it out, visit the site below on a desktop browser (I haven’t had time to do mobile adaptation yet). Register an account and get 5 free credits. There are also onboarding tasks and referral rewards for earning more credits.

https://mycareerhelp.ai

Here’s the tech stack I used:

• Backend: Python, Django, MySQL, Redis, Celery, Gunicorn, Nginx
• Frontend: React, Chakra UI
• Cloud services: DigitalOcean, Cloudflare
• Email services: Resend + Zoho

Managing My Own Expectations

Honestly, from a business perspective, this direction isn’t ideal. The market is saturated with low barriers to entry, and there’s not much money in it. From a product standpoint, it’s simple — barely worth turning into a full product. You could get similar results with a chatbot and a well-crafted prompt. Don’t believe me? I built the prototype in 30 minutes using Dify. I just structured the input and output, added some business-oriented prompts, and workflows. Essentially, it’s a repackaged tool with a bit more polish.

Platforms like Maimai, Boss Zhipin, or LinkedIn are better positioned for this since they already have the data, scenarios, and user habits in place. (For example, Maimai charges 369 RMB — about $52 — just for a single resume diagnosis and customization.)

So, if you ask what kind of competitive advantage I have, I don’t really have one. But this isn’t about following the big company playbook. I don’t need a project proposal, market analysis, competitive assessment, or cost analysis to justify the resources. I just need to decide, take action, and own the results. The decision, the action, and the outcome are all mine.

But, there is indeed one clear difference: platforms like Boss Zhipin are designed to capture user information and sell it to B2B clients (over 99% of Boss Zhipin’s nearly 6 billion RMB revenue in 2023 came from B2B online recruitment services, with less than 1% from individual users). I just want to build a simple tool. All users need is a verified email to get started. They can even replace their school or company names with placeholders like XYZ. There’s no need for real names or phone numbers, and even if personal info is uploaded, I strip it out during parsing because my database doesn’t store it. If users want to leave, they can delete their account with one click, and all their data will be erased.

Of course, one could argue I’m limited to building a tool because I lack the resources to create a full platform. This does limit my potential. Platforms offer more room for growth and engagement than standalone tools. From a user perspective, tools tend to be “use it and leave.” If users don’t come back, the business value won’t grow.

• Is this path viable? Maybe. Hard to say.
• Is there a lot of room for growth? Not really. Most job seekers are blue-collar, and only a small percentage of white-collar workers need resume optimization and career advices.
• What if a big company comes in and does it? No need to worry — it’s already happening, but the pie is too small for them to care much.
• Will it be profitable? Hard to say, probably not, but at least the costs are low for now.
• So why do it?

The truth is, I could do nothing. In this economy, having a stable job at a big company — trading time for security — is already a solid choice. Or I could just lay low, become a low-consumption slacker, and enjoy life on a beach somewhere. As long as I’m not raising kids, there’s not much to worry about. But I can’t settle. At this stage of life, with the ability, health, and capital to experiment, it feels wrong to coast. Who knows — years from now, will I regret doing something, or regret not doing something?

I still want to do something. There’s always that first step, testing the waters to see if they’re deep or shallow. If this path doesn’t work, I’ll pivot. It’s not about one product or project — it’s about the journey of independent development. Even if it turns out to be a pipe dream, at least I tried.

In Conclusion

I once read a quote: “If you’re not ashamed of your first version, then you probably launched too late.” It’s true. I should have launched in early August, but I worried it would be labeled as “nothing special.” I feared that, at my age, what I built would still seem basic, and it would shatter my so-called “image.” (Of course, there were also factors beyond my control that delayed the launch, like registering the company and opening a bank account — things I’d never dealt with before, but at least I gained some new experience points.)

But during a typhoon day, as I lay on the couch petting my cats, I realized something important: I AM “nothing special.” I’m just someone transitioning from a product manager to an independent developer. Like when I prompt AI, I always add, “I’m a new learner in [specific field].” There’s no shame in that. While my brain still works, I’ll use AI to do what I want and become who I want to be.

I AM a new learner, and I’m ready to explore this new world.