How I Write

Every article on this site goes through a structured pipeline before it reaches you. I am transparent about this process because I think the industry needs more honesty about how content gets made, especially when AI is involved.

The Short Version

I use AI as a research and drafting collaborator. Every idea, opinion, and editorial decision is mine. The AI does not decide what I write about, what position I take, or whether an article is good enough to publish. I do.

Research

Every article starts with a topic I have been thinking about, usually something I encountered at work, read in a paper, or debated with a colleague. The research phase is methodical:

1. Deep sourcing: I collect primary sources first: regulatory text, academic papers, company engineering blogs, official documentation. Secondary sources (analyst reports, expert commentary) fill gaps. Vendor marketing and unsourced opinion pieces are flagged and used sparingly, if at all.
2. Source quality scoring: Every source is classified as Primary, Secondary, or Tertiary. If an article has fewer than two Primary sources or more than 40% Tertiary sources, it goes back to the research phase.
3. Source encapsulation: Each source gets a structured summary: title, URL, author, date, key claims, data points, and direct quotes. This creates a paper trail I can cross-reference later during fact-checking.

I am looking for the ground truth, not the consensus take. If every article on a topic says the same thing, that is usually a sign that nobody did original research and everyone is citing each other. I try to find the primary data and form my own view.

AI Collaboration

Here is where I expect skepticism, and I welcome it. I use an AI agent (Claude) throughout the writing process. But "use AI" is vague to the point of being meaningless. Here is exactly what it does and does not do:

What AI does

• Research acceleration: Fetching, reading, and summarizing sources across dozens of URLs in minutes instead of hours.
• First draft generation: Producing a structured first draft based on my research notes, outline, and specific instructions about angle and tone.
• Quality checks: Running automated scans for factual accuracy, broken links, readability metrics, and voice authenticity.
• Mechanical editing: Fixing formatting, checking reading time calculations, validating that discipline names use consistent capitalization.

What AI does not do

• Choose topics: Every article originates from my professional experience or intellectual curiosity. The AI does not suggest what I should write about.
• Form opinions: The positions, arguments, and conclusions in every article are mine. If I disagree with what the AI drafted, I rewrite it until it reflects what I actually think.
• Approve publication: No article publishes without my explicit approval. I review every piece on my phone first, often multiple times, and collect feedback before deciding it is ready.
• Replace domain expertise: I have spent two decades in data, AI, and platform engineering. The AI helps me write faster. It does not supply the judgment that comes from having built and broken these systems firsthand.

The best analogy I have: the AI is a very fast research assistant who can also type. I am the editor-in-chief. The assistant does not set the editorial direction, and it does not get a vote on whether something ships.

Editorial Standards

After the first draft exists, the real work begins. This is where most AI-assisted content falls apart, and where I spend the majority of my time.

1. Preview review: Every draft deploys to a private preview URL. I read it on my phone first, often multiple times. Reading on a different device, in a different context, catches problems that reading in an editor does not.
2. Feedback collection: Draft articles go through a private preview phase where I collect feedback from trusted readers before publishing. That feedback directly shapes the final version. Not all of it gets implemented, but all of it gets considered.
3. Voice check: Does this sound like something I would actually say? I maintain a list of phrases that no article is allowed to contain, because they are reliable markers of writing that prioritizes sounding smart over being clear. If a paragraph reads like a corporate whitepaper or a generic blog post, it gets rewritten or cut. Every article must contain at least one personal experience from my own work. This is a hard rule, not a suggestion.
4. Fact verification: Every factual claim gets cross-referenced against its source. Not "does this sound right?" but "does the source actually say this, with this number, on this date?" Claims that cannot be verified get removed or softened.
5. Originality check: Every article is checked for originality against published web content before it goes live. If passages are too close to existing writing, they get rewritten or properly attributed. Nothing publishes below an originality threshold.
6. The final question: Does this article reflect my genuine understanding of the topic? If I were sitting across from a senior colleague and they challenged any point in this piece, could I defend it from experience? If the answer is no, the article is not ready.

Beyond editorial judgment, mechanical checks run on every article: readability metrics, link validation, source quality scoring, and reading time verification. Automating these frees editorial energy for the judgment calls that actually matter: whether the argument holds, whether the voice is authentic, and whether the piece is worth publishing at all.

Why This Matters

The internet is filling up with content that was generated in seconds and published without scrutiny. I do not want to add to that pile. AI makes me faster at the parts of writing that benefit from speed (research, drafting, mechanical checks) so I can spend more time on the parts that require a human (forming opinions, evaluating arguments, deciding what is worth saying at all).

If you read something on this site that strikes you as wrong, unsubstantiated, or not reflective of real-world experience, I want to hear about it. Every article has my name on it, and that means something to me.

Curious about the technical infrastructure? Read How This Site Is Built.