Nine Products, One Person, Zero Hype: What AI Actually Changed About Building Solo
People ask me how I run nine products alone.
The honest answer is: I don’t. I run them with an AI agent that does the work of three junior developers, one designer, and a content writer who never sleeps. But even that framing is misleading, because it makes it sound seamless. It isn’t.
Let me break down what’s actually happening, what it costs, and what AI genuinely changed versus what it didn’t.
The stack: what I actually pay for
Most “solo developer stack” articles list tools I’ve never heard of and would never use. Here’s what’s actually running, with real numbers.
Infrastructure: $0/month
Everything runs on Cloudflare’s free tier. Nine Cloudflare Pages projects, three D1 databases, four KV namespaces, two R2 buckets, and a handful of Workers. The free tier gives me 100,000 Worker requests per day, 500 builds per month, unlimited bandwidth on static assets, and 10GB of R2 storage.
I have never hit a limit. Not once. The sites I run are content-heavy, not traffic-heavy. If any of them ever outgrow the free tier, that’s a problem I want to have.
AI tools: ~$30/month
This is where it gets interesting. My AI agent runs on GLM (zhipu), not GPT-4. The reason is simple: GLM costs a fraction of what GPT-4 costs, handles Chinese and English equally well, and is fast enough for real-time coding. I pay roughly $30/month in API calls across all nine products.
For comparison: Uber blew through its entire 2026 AI budget in four months. Microsoft pulled Claude Code licenses from developers after costs spiraled. One company found a $500 million Claude bill because nobody set usage limits. I spend $30. The difference isn’t that I’m smarter. It’s that my use case (content generation, code review, SEO analysis) doesn’t require frontier-model reasoning on every call.
Domains: ~$120/year
Nine .xyz, .ink, .dev, and .si domains. Roughly $13 each per year. This is the most boring line item and also the most essential.
Payment processing: $0 fixed
Polar takes a percentage per transaction. Creem did too, before they shut down our account for selling “services” (which their policy technically prohibits). The lesson: never build your business on a platform whose terms can change overnight. We moved to Polar and never looked back.
Total monthly cost: ~$40.
A senior developer costs $10,000-15,000/month fully loaded. My entire operation costs less than a Netflix subscription.
What AI actually changed
Everyone talks about AI coding tools. That’s the wrong frame. AI didn’t make me code faster. It changed what I spend time on.
Before AI (2021-2023): I spent 80% of my time writing code and 20% thinking about what to build. The bottleneck was execution speed. I had more ideas than hours.
After AI (2024-2026): I spend 20% of my time reviewing code my agent wrote and 80% on strategy, research, and distribution. The bottleneck shifted from “can I build it” to “should I build it, and will anyone find it.”
This is the change that matters. Not the typing speed. The reallocation of cognitive bandwidth.
Code generation: real, but overhyped
My agent writes competent code. Astro components, Python scripts, SQL migrations, CSS. Probably 70% of the codebase across nine products was written or substantially drafted by AI.
But here’s what the productivity gurus don’t tell you: reviewing AI code is slower than writing it yourself, if you care about quality. AI generates plausible code that looks right and sometimes isn’t. The amount of time I spend debugging subtle issues, logic errors where each line is individually correct but the overall flow is wrong, and CSS that works on desktop but breaks on mobile is significant.
The net gain is real but not 10x. It’s maybe 2-3x for code, and only because I’ve learned to write very precise prompts and I have a deep enough understanding to catch mistakes quickly.
Content creation: genuinely transformative
This is where AI delivered the most value. OnchainDiary has 240+ educational articles. Echoir has 200+ SEO pages. ZensInk has documentation, changelogs, and marketing copy. LiteInk has a media publication with original fiction.
Writing all of that by hand would have taken years. With AI, it took months. And the quality is higher than what I could produce alone, because the AI forces structure: every article has a BLUF block, internal links, FAQ schema, and consistent formatting that I would never maintain consistently across hundreds of pages.
SEO and research: the hidden multiplier
ZensInk exists because I needed SEO tools and couldn’t afford Ahrefs. The irony is that ZensInk itself became a product. The SEO workflow, keyword research, rank tracking, content gap analysis, all automated through AI agents reading data and writing reports, is the single most valuable thing AI does for my business.
It’s not glamorous. Nobody retweets “I automated my SERP analysis.” But it’s what allows nine products to compete for search traffic against teams with dedicated SEO departments.
The part nobody talks about: distribution
AI made building free. Distribution didn’t get cheaper.
Here’s the uncomfortable truth: I can build a product in a weekend. Getting someone to use it takes months. The cost of creating software dropped to near-zero. The cost of reaching people who want that software stayed exactly the same.
This is why most solo builders fail. Not because they can’t build. Because they build first and think about distribution second, if at all.
My distribution strategy across nine products:
- SEO (long-term, 6-12 months to see results): OnchainDiary, ZensInk, Echoir
- Social proof (GitHub stars, community): ZensInk open-source, Ink CMS
- Direct sales (marketplace listings): LiteInk templates on Polar, ZensInk on Xianyu
- Content marketing: LiteInk Media (AI analysis + fiction), uzenlabs blog
Each channel takes consistent effort over months before it compounds. AI can generate the content, but it can’t build the relationships, earn the backlinks, or create the community trust that makes distribution work.
What doesn’t work
I’d be dishonest if I only shared the wins. Here’s what AI didn’t fix.
Context switching kills productivity. Nine products means nine codebases, nine design systems, nine deployment pipelines, nine sets of problems. Even with an AI agent that can switch contexts instantly, I can’t. Every time I context-switch, I lose 30 minutes of ramp-up time. On a good day, I work on two products. On a bad day, I spend the whole day context-switching and ship nothing.
Some problems require human judgment. Pricing strategy. Brand positioning. Which feature to cut. Whether a market exists before you build the product. AI can analyze data and offer options, but the final call is always mine, and it’s usually the part that matters most.
Maintenance is the hidden tax. Nine products need updates. Dependencies break. APIs change. DNS records expire. Cloudflare changes their free tier limits. A page that ranked #3 drops to page 2 and you have to figure out why. This is unglamorous work that doesn’t show up in any productivity dashboard but eats 30% of my time.
The numbers that actually matter
Let me end with what I think matters most, and what I wish someone had told me two years ago.
How many products make money? Two. ZensInk (SEO tool sales + done-for-you services) and LiteInk (Astro template sales). The other seven are either free, archived, or pre-revenue.
How many users do I have? Across all products: maybe 3,000 monthly active users. This sounds small. It is small. But it’s enough to validate that the products work and to generate enough revenue to cover costs.
How much do I work? Roughly 10-12 hours a day, six days a week. AI didn’t reduce my hours. It increased what I produce in those hours.
Would I recommend this path? Only if you genuinely enjoy building things. Not the idea of building things. The actual, unglamorous, debugging-CSS-at-midnight reality of it. AI didn’t make solo building easy. It made it possible. Those are different things.
The hype says AI will create a million one-person unicorns. The reality is that AI lets one person do what used to require three. That’s not a unicorn. That’s a lean, sustainable, independent operation. And for me, that’s enough.
I’m Jask, an independent researcher and developer. I run UZEN Labs. This is a build log, not a pitch deck.