How to Write a Blog (or: What I Learned By Doing It)
I dislike the word “blog.” I want to get that out of the way immediately, because it is the first thing I thought when someone suggested I start one: the word is ugly. It sounds like something that happens to a drain. “Web log” is somehow worse, conjuring the image of a ship captain who learned HTML. I avoided starting one for years partly on the strength of that aesthetic objection, which I now recognise as displacement behaviour for a deeper discomfort, the kind that asks: who do you think you are, putting your opinions on the internet?
That question turned out to be more useful than the word problem. It forced me to answer some things I had been avoiding, and those answers shaped everything that came after.
The questions you need to answer before you write anything
Before the first post, I had to sit with a few questions that felt trivially obvious but turned out not to be. Why am I doing this? Who is it for? What would success look like, and, less comfortably, what would failure look like?
On the why: I write because I think better when I write. The process of turning a loose opinion into a paragraph that has to hold up under its own weight forces a kind of intellectual honesty that thinking alone does not. If I cannot explain something in prose, I probably do not understand it as well as I thought. That answer is sufficient on its own, and it does not require an audience.
On the who: the audience I write for is primarily myself, and secondarily anyone who finds the post useful. That ordering matters. Writing for a hypothetical reader who needs to be entertained or impressed produces a different kind of prose than writing to clarify your own thinking and then leaving the door open. The former produces performance; the latter, occasionally, produces something worth reading.
On success and failure: I do not measure success in page views. I measure it in whether the post says something I could not have said without the discipline of writing it, and whether I am willing to have my name on it in ten years. Failure, by that standard, is publishing something I do not believe or padding a post to a length it does not earn. Those criteria are entirely within my control, which is another way of saying that my success criteria do not depend on the internet agreeing with me.
Process: reading first, writing second
The most reliable piece of writing advice I have encountered, across academic writing, fiction, journalism, and personal essays, is this: writing is a downstream activity. The quality of what comes out is almost entirely determined by the quality of what went in. The ratio I have settled on, only half-jokingly, is a hundred parts reading to one part writing, and I think it applies regardless of whether you are writing a PhD thesis, a fantasy novel, or a post about audiobooks.
This is not a prescription to read voraciously before every post, though that helps. It is more that the instinct to sit down and write before you have genuinely absorbed the subject tends to produce prose that sounds thin in ways the writer cannot identify but readers immediately feel. Opinion writing that has not been stress-tested against counterarguments, or technical writing that has not been informed by people who know more than you, tends to produce the same symptom: confident sentences that do not actually know what they are talking about.
My process is approximate and not especially disciplined. I read around a topic until I have an opinion that feels load-bearing, meaning an opinion I can support with something more than vibes, and then I write a draft that is usually too long and too hedged, and then I cut it back. The cutting is where most of the work happens. A first draft is an argument trying to find its shape; a final draft is the same argument with most of the self-doubt removed.
What I do not do is outline first. I tried it and found that outlining before I had written anything just produced a map of things I already thought, which was less interesting than finding out what I thought in the act of writing. Your mileage will vary on this.
Dealing with criticism, including the unfair kind
I wrote a post about Patrick Rothfuss and The Doors of Stone that, when I shared it, attracted a number of comments asserting it was AI-generated and that the quality was poor. I want to be honest about what that felt like: it stings, even when you know the criticism is wrong, and even when you are not especially invested in the opinion of the person delivering it. That sting is real and worth acknowledging rather than performing indifference about.
What I found, though, is that criticism worth receiving tends to surface within the noise, and criticism not worth receiving can teach you something about the limits of your audience, or occasionally about your own prose. That Rothfuss post had sections I was uncertain about. The criticism did not identify those sections, but it prompted me to reread the post with fresh suspicion, and I found a few places where I had leaned on hedging language rather than committing to a point. The criticism was wrong about the diagnosis and accidentally useful about the symptom.
The harder thing to accept is that you cannot force people to read your work carefully. You also cannot force friends and family to be interested, and trying tends to produce the opposite of the intended effect. The people who matter will find the posts they find useful; the people who do not care will not, and pushing your writing at them builds resentment faster than it builds readership. A blog written for yourself does not need a marketing strategy. It needs to be good enough that the people who do find it are glad they did.
If someone accuses you of using AI to write your work, and you did not, the only real response is to keep writing in a way that proves otherwise over time. Trying to argue the point in the comment section is undignified and usually futile. One thing worth keeping in mind: the accusation of AI use has become a generic criticism, deployed the way “formulaic” used to be deployed, meaning it often says more about the reader’s expectations than the writer’s process. Take it as data, not as verdict.
What to write about
This is the question I get asked most often, and my answer is not very useful: write about things you have thought about long enough to have something to say. That sounds obvious until you try it, at which point you discover that most topics you feel confident about either have obvious correct answers or are genuinely contested in ways that make your opinion one among many with no particular claim to attention. Both of those realisations are useful.
On facts: if you are writing about something where facts are in play, cite people who know more than you. Not out of academic formality, but because the alternative is to ask readers to trust your memory and judgement on things that can be checked. I try to link to sources or name the people I am drawing on wherever I make a factual claim. I am not always as diligent about this as I should be, but the instinct is right.
On topics: I have no formal rules about what I cover. I write about books, data science, where I grew up, audiobooks, AI ethics, and anything else I have accumulated enough thought about to fill a post. The only filter I apply is one I call the four-audience test, borrowed from the idea of writing something you would be comfortable having read by a friend, a foe, a family member, and your boss. If a post fails that test, it either needs to be rewritten or left in the drafts folder. That is not a rule about being inoffensive: it is a rule about being honest enough about your own motives that you would not regret the post existing in a world where all four of those people read it on the same day.
Resources and tools
A short list of things I have found genuinely useful, rather than a comprehensive guide to tools that exist:
On writing itself: William Zinsser’s On Writing Well is the book I recommend most often, because it addresses the problems writers at every level actually have: prose that is too long, too hedged, too fond of its own vocabulary. Joseph Williams’ Style: Lessons in Clarity and Grace is more systematic and more demanding, and worth reading if Zinsser leaves you wanting a framework rather than principles.
On AI as a writing tool: I use local language models for review rather than composition, and I think that distinction matters. An AI reviewer will catch structural problems, identify where hedging language has crept in, and flag sentences that are technically grammatical but doing no useful work, which are things worth catching. An AI writer produces prose that is grammatically fine and intellectually shallow: it will fill a page with sentences that sound like they are saying something without committing to a position. The insight density of AI-written text is low in a way that is difficult to locate sentence by sentence but immediately apparent across a whole post.
I wrote about this more directly in my post on building a second brain with local LLMs: the short version is that a local model running on your own hardware is preferable to a cloud service for review work, both for privacy and because it changes how you interact with the tool. Chatting through a draft with a local model, asking it to challenge your assumptions or identify where you have asserted something without supporting it, is a different activity to asking it to write the post for you. The former produces better writing; the latter produces a post that will read like a post that asked an AI to write it, which is a recognisable and increasingly unflattering register.
On where to blog: I will write a separate post about hosting a personal website and blog, covering the technical options from the simplest to the most involved. For now: the platform choice matters less than whether you control the content. Substack, Medium, and similar services offer distribution but extract a degree of control over presentation and, in some cases, discoverability.
On structure and components: I keep a template reference document for the reusable patterns I have built into this blog, covering things like photo captions, expandable FAQ sections, timelines, and glossary boxes. If you are building your own static site, a similar document saves considerable time when writing the tenth post.
Privacy, anonymity, and platform
Blogging is putting yourself on the record. That is worth taking seriously before you start, not to discourage you from starting, but because the decisions you make early about what you share and under what name are difficult to undo. I write under my own name because I am writing about things I would stand behind in any other context, but that is a choice that carries tradeoffs: professional visibility, searchability, and the permanent internet archive of your younger opinions all come with it.
Anonymity is a legitimate option, especially if you are writing about topics where your professional identity and your personal opinions are in tension. The tradeoff is that anonymous writing carries less authority: readers cannot evaluate your credibility the way they can when you attach a name and a record to a claim. That may not matter for what you are writing, but it is worth knowing.
On platform: writing on your own website is, by internet standards, a niche activity, and you should go in knowing that the audience for a personal site built on a custom domain and hosted on GitHub Pages is not going to grow the way a Substack with SEO-optimised titles might. That is fine, and may even be preferable, depending on what you are writing for. I find the small audience clarifying: nobody is reading this because an algorithm pushed it at them, so the people who do find it tend to be the people it was written for.
The audience question brings me back to where I started. Writing for yourself, seriously and honestly, with enough care that what you produce is worth someone else’s time: that is the whole of what this kind of writing is. You cannot force anyone to read it. The word is still ugly. Do it anyway.