Five months ago, this blog didn’t exist. I had a domain name, a vague idea about writing developer content, and a theory that AI could shortcut the hardest parts of building a content site. Today, the site gets over 10,000 unique visitors per month — and it’s still growing.
This is the full story. What I built, how I used AI at every step, what actually moved the needle, and what was a complete waste of time. If you’re thinking about starting a developer blog in 2026, this might save you a few months of trial and error.
Starting from zero
I wrote about the initial setup in How I Built a Passive Income Blog with AI in One Day. The short version: I picked Astro as the static site generator, deployed on Vercel, and wrote everything in Markdown. Total infrastructure cost: basically nothing. A $12 domain and free hosting.
That tech stack turned out to be one of the best decisions I made. Astro generates pure static HTML, which means pages load fast, Google can crawl everything easily, and I never have to worry about server costs scaling with traffic. When you’re trying to grow from zero, you need every technical advantage you can get.
But having a fast, well-structured site means nothing without content. And that’s where things got interesting.
The content strategy that actually worked
I didn’t start with a content strategy. I started by publishing whatever seemed useful and watching what happened. After about six weeks and maybe 40 posts, patterns started to emerge in my analytics. Some content types consistently outperformed others — by a huge margin.
Content clusters and pillar pages
The biggest unlock was thinking in clusters instead of individual posts. Instead of writing one article about Ollama and moving on, I built an entire cluster: a cheat sheet, a connection error fix, a guide to running specific models locally, comparison posts against alternatives. Each post links to the others, and they all link back to a central pillar page.
Google seems to love this. When you have 8–12 posts all covering different angles of the same topic, with strong internal linking between them, individual posts start ranking faster than standalone articles. My Ollama cluster went from zero to 2,000+ monthly visitors within about two months.
The three content types that drive traffic
After months of data, three types of posts consistently bring in the most organic traffic:
1. Error fix posts. Things like “ENOSPC file watchers fix” or “Cannot read properties of null fix.” These target developers who are actively searching for a solution to a specific problem. The search intent is crystal clear, competition is often low, and the posts are straightforward to write. They’re not glamorous, but they’re traffic machines.
2. Cheat sheets. Concise reference pages for tools and frameworks — PostgreSQL, Nginx, Ollama, Git. Developers bookmark these and come back repeatedly. The return visitor rate on cheat sheets is 3–4x higher than on regular articles. They also earn backlinks naturally because people share them in team Slack channels and GitHub discussions.
3. Comparison and “best of” articles. Posts like Best AI Coding Tools in 2026 and The Real Cost of AI Coding Tools pull in readers who are in research mode. These take longer to write and need regular updates, but they rank for high-value keywords and tend to convert well for affiliate links.
What didn’t work: generic tutorial content. “How to use Python decorators” or “Introduction to Docker” — topics where there are already thousands of well-established articles. I published a handful of these early on and they went nowhere. The internet doesn’t need another beginner Docker tutorial.
How I used AI at every stage
AI wasn’t just a writing tool — it was involved in almost every part of the workflow.
Writing with Claude
I use Claude for first drafts of almost every post. But the process isn’t “prompt and publish.” I research the topic manually, write a detailed outline, have Claude draft based on my notes, then rewrite heavily — cutting fluff, adding personal experience, fixing anything generic. Final pass is for SEO: headings, meta description, internal links.
AI saves me 40–50% of writing time per post. But editing is non-negotiable. Raw AI output reads like raw AI output, and readers (and Google) can tell. The posts that perform best are the ones where I added the most of my own voice.
Scripts for content generation
For structured content like cheat sheets and comparison tables, I built simple scripts that call the Anthropic API to generate formatted Markdown sections. For example, I have a script that takes a CLI tool name and generates a cheat sheet skeleton with the most common commands, grouped by category. I then review, test every command, and fill in the gaps.
This is where AI shines brightest — not in writing prose, but in generating structured, factual content that I can verify and polish.
IndexNow for faster indexing
I set up IndexNow to ping search engines every time I publish or update a post. Astro makes this easy since you know exactly which URLs changed at build time. A small post-build script sends updated URLs to the IndexNow API.
Before IndexNow, new posts took a week or more to appear in Google. After: 24–48 hours. When you’re publishing frequently, that difference compounds fast.
The SEO approach
I’m not an SEO expert. I didn’t hire one either. My approach was simple and mostly learned by doing.
Long-tail keywords
I almost never target broad, competitive keywords. Instead of trying to rank for “AI coding tools,” I target things like “Claude Opus 4 vs GPT-5 for coding” or “how to fix Ollama connection refused error.” These long-tail queries have lower search volume individually, but they’re much easier to rank for, and they add up.
I use a combination of Google Search Console data (to see what queries are already bringing impressions) and common sense (what would I search for if I had this problem?). No expensive SEO tools.
The Chinese AI model niche
This was probably my biggest traffic win, and it happened partly by accident. I started writing about Chinese AI models — Qwen, DeepSeek, Kimi — because I was genuinely curious about them and nobody else in the English-language developer blogging space was covering them well. The timing was perfect. Interest in these models exploded, and my posts were among the few English-language resources available.
The lesson: find topics where demand is growing but supply is still thin. That’s where a small blog can compete with established sites.
Internal linking
Every post links to at least 2–3 other posts on the site. Each cluster has a clear linking structure where related posts reference each other. This helps Google understand topical authority and keeps readers moving between pages.
My average pages-per-session went from 1.2 to 1.8 after I added internal links to older posts. That’s a meaningful difference.
What didn’t work
Not everything was a win. Some honest failures:
Publishing too many articles at once. Early on, I had a week where I published 15 posts in three days. I thought more content = more traffic, faster. Instead, Google seemed to treat it as suspicious. Several of those posts took much longer to index than posts I published at a steady pace of 1–2 per day. I can’t prove causation, but I’ve stuck to a more consistent publishing schedule since then and the results have been better.
Generic content without a unique angle. Any post where I didn’t have personal experience or a specific opinion to add performed poorly. The posts that rank are the ones where I tested something myself, hit a real problem, or had a genuine take. AI can write generic content perfectly well — which is exactly why generic content doesn’t stand out anymore.
Ignoring post updates. Some of my early comparison posts went stale within weeks because AI tools change so fast. I lost rankings on a few posts before I realized I needed to build updating into my workflow. Now I review and refresh high-traffic posts at least monthly.
Monetization
I’ll keep this brief since the focus here is traffic, not revenue. The site is monetized through affiliate links — mostly for AI tools, hosting services, and developer products. I won’t share exact numbers, but the revenue is meaningful enough to justify the time, and it’s growing roughly in proportion to traffic.
The key insight: comparison posts and “best of” lists convert far better than tutorials or error fixes. Someone reading “best AI coding tools” is in buying mode. Someone reading “how to fix a broken pipe error” just wants their code to work.
The numbers
Here’s the rough traffic trajectory:
| Month | Unique visitors |
|---|---|
| Month 1 (March) | ~200 |
| Month 2 (April) | ~800 |
| Month 3 (May) | ~2,500 |
| Month 4 (June) | ~6,000 |
| Month 5 (July) | ~10,000+ |
The growth wasn’t linear. Months 1–2 were painfully slow. Most of the early traffic came from a handful of error fix posts that happened to rank quickly. The real inflection point came in month 3 when the content clusters started kicking in and Google began treating the site as a topical authority in certain niches.
What I’d do differently
If I started over tomorrow:
- Start with clusters, not random posts. I wasted my first few weeks publishing scattered content. I’d pick 3–4 topic clusters from day one and build them out systematically.
- Set up IndexNow immediately. I didn’t add this until month 2 and lost indexing speed on dozens of early posts.
- Skip the generic content entirely. Every generic tutorial I wrote was a waste of time. I’d focus exclusively on error fixes, cheat sheets, comparisons, and opinion pieces from the start.
- Build an email list earlier. I still don’t have one, and I regret it. Organic traffic is great but fragile — one algorithm change could cut it in half overnight.
Lessons learned
Building a blog with AI isn’t about having AI write your content for you. It’s about using AI to remove friction from every step — research, drafting, formatting, structured data — so you can focus on what differentiates your content: personal experience, genuine opinions, and underserved niches.
The 10K milestone is just a number. What matters more is that I’ve built a system that compounds. Every new post strengthens the clusters, improves internal linking, and gives Google more reasons to trust the site.
If you’re thinking about starting something similar: just start. The tech stack is free. The AI tools are accessible. The only thing you can’t shortcut is showing up and publishing consistently.