Medium: AI Engineering Earns $5,000 in a Month

A Record Month on Medium

In May, my earnings on Medium surpassed $5,000, an impressive figure achieved through my writings on AI engineering. This success is all the more remarkable considering that the previous month, the same account recorded 10.9K views, and two months prior, only 3.9K. This spectacular leap might seem like the result of a few viral articles, but the reality is more complex. Indeed, if these articles had been published on a brand new account, they would likely have experienced a spike before tapering off. However, on my account, they sparked interest in about fifteen other articles. Readers, after reading an article on local LLMs, explored my profile and clicked on other articles about augmented generation through retrieval and agent architectures.

A Gradual but Significant Change

March had been relatively quiet. I was publishing regularly, but signs of engagement were sporadic: a comment here, a highlighted code snippet there. Traffic was unpredictable. However, by mid-April, a change began to manifest. Even without immediate viral success, my baseline daily views were increasing. Each new article published triggered a secondary wave of traffic to older ones. Thus, the catalog came to life, and readers discovering one of my works began to explore the others.

In May, the situation evolved significantly. Recommendations ceased to be random, and a construction-focused type of reader began to appear regularly. When an article took off, traffic spread throughout the rest of the catalog. I remember one Tuesday morning when, refreshing the dashboard, I saw an old article on RAG from March suddenly reappear in the daily rankings, right next to an article I had published twelve hours earlier.

Not Relying Solely on Viral Articles

When people see such a fruitful month, they often seek to understand which articles generated the most success. Looking at my May dashboard, it is clear that a few specific articles generated the majority of the numbers. My guide on setting up Claude Code attracted 39,000 views and nearly 20,000 reads, bringing in about $1,360. Another article on the best local LLMs for programming also garnered 39,000 views and 18,900 reads, generating just over $2,000—it was my highest-earning article. A list of 12 AI books brought in 28,000 views, 12,800 reads, and $714.

It would be tempting to conclude that I should simply write more lists and setup guides. However, these spikes only mattered because there were 20 other articles for readers to land on afterward. When a developer clicked on the setup guide for Claude Code, they found a helpful tutorial. At the bottom of that page, they saw links to my other works. They discovered in-depth technical dives like "RAG Chunking That Works," "Multi-Agent Systems," and a detailed comparison of LangGraph versus Temporal. If my profile had only contained generic news about AI, that developer would have closed the tab. Because it contained dense engineering articles, they realized I was a builder. They clicked the follow button, joined the mailing list, and checked out older articles.

A Well-Defined Niche

I drew a strict boundary around my topics. The niche was not artificial intelligence in general, but specifically AI engineering. If I wrote about AI broadly, I would be competing with news sites and generic content farms. I would end up writing about how AI will change the future of work, which no engineer wants to read. Instead, I broke down my niche into very specific pathways. This specific mapping worked. It had enough search volume to be relevant, but the code snippets naturally filtered out people looking for prompt tricks for ChatGPT. It was also deeply connected. If someone reads an article on RAG chunking, they are naturally interested in an article on reranking or hybrid search. I had to remain disciplined here. Writing an excellent article on Python agents today and a generic article on productivity tomorrow would simply dilute the exact reader base I was trying to build.

Composition: A Slow but Effective Process

By the end of April, I realized that composition in content creation resembles absolutely nothing for weeks. It feels like tedious little movements. I would publish an article and get 50 views. I would publish another and get 70. Then, the older articles began to trigger together. The pattern repeated every time. A new article attracted attention. Readers clicked on it, finished reading, and the recommendation engine would bring up an article from three weeks ago at the bottom of the page. The spike from Claude Code directly helped my older articles on AI coding workflows. The article on AI books redirected traffic to my roadmap articles. The article on RAG Chunking maintained my technical authority on RAG and fueled traffic to my reranking guide. You cannot judge a technical article by its traffic in the first week, as an article on Python agents may remain dormant for a month until a related article on LangGraph suddenly wakes it up.

The Importance of Technical Dives

I previously thought it was necessary to choose: in-depth tutorials or broad overviews. May taught me that both are needed, for different reasons. In-depth technical dives proved my competence. My article on RAG Chunking was dense, filled with regex patterns for markdown parsing and detailed explanations of semantic limits. It did not go viral. It recorded 2,700 views and 1,100 reads. But those who clicked on it read it, with a reading rate of about 40%. Meanwhile, my in-depth architectural comparison of LangGraph versus Temporal only garnered 959 views but brought in $119.91. That's a high earnings-per-view ratio. It monetized well because it addressed a specific architectural question that senior engineers were searching for. They were trying to decide between state machines and durable execution for their agent orchestration, and they spent ten minutes reading every word of that article. The biggest discovery came from articles with broader entry points.