Writing Advice Is Mostly Just Personal Preference

Much of what passes for writing advice is really just a description of how the author writes, dressed up as universal guidance. The implication is always "you should write like me," which is no more useful for writing than it is for technical work. A style that works for one person in one context often fails elsewhere. Different goals, different constraints, and different strengths all call for different approaches. Copying another writer's method wholesale is unlikely to serve you well.

The range of what can succeed is wide. Among the most-read programming bloggers from roughly 2000 to 2017 were Joel Spolsky, Paul Graham, Steve Yegge, and Julia Evans — four writers whose styles differ dramatically along almost every axis: topic choice, prose style, post length, humor, technical depth, evidence use, and nuance.

Length alone demonstrates the point. Julia Evans and the author both started blogging in earnest in September 2013. Over the years, they produced roughly the same total word count — but she published about an order of magnitude more posts. Both approaches worked fine.

Consider the difference in rhetorical stance. Paul Graham's writing exemplifies what Thomas and Turner call "classic style," as described in Clear and Simple as the Truth. The prose is clean and direct, and more importantly, the presentation itself is treated as sufficient — no supporting evidence is needed because "truth needs no argument but only accurate presentation." Rochefoucauld's sketch of Madame de Chevreuse embodies this ideal:

Madame de Chevreuse had sparkling intelligence, ambition, and beauty in plenty; she was flirtatious, lively, bold, enterprising; she used all her charms to push her projects to success, and she almost always brought disaster to those she encountered on her way.

The writer poses as having complete knowledge, presenting conclusions as timeless absolutes, like a geometric proof. All personal history and reasoning vanish from the expression.

The opposite pole is the author's own style, which is characterized by long, meandering sentences that mirror how he thinks. He prefers structured arguments backed by evidence, with caveats about where evidence doesn't apply. Omitting supporting evidence may read more cleanly, but it forces the reader to accept or reject bare assertions. Understanding the underlying reasoning allows readers to agree or disagree on substance rather than trust.

Fit Matters More Than Popularity

There is no single right answer on length, style, or any other dimension. If raw popularity is the goal, publishing short posts frequently appears more effective than infrequent long ones. Confident, unqualified statements tend to draw readers because people value confidence. But one can be plenty popular without optimizing for it, and other goals may matter more.

Some topics simply cannot be covered in 500 or 1000 words. The author's post on branch prediction, for example, runs 5.8k words and pursues two goals: giving a programmer with no architecture background a historical survey, and teaching them enough to read a modern branch prediction paper. Every predictor discussed is a necessary building block for later concepts, or historically important. A shorter version explaining that branch prediction exists would reach a wider audience — but it wouldn't be the post he wanted to write.

Style constraints follow similar logic. Writing about big ideas at a high level is amenable to a clean, elegant style. But discussing low-level technical details in depth often requires messier prose. The long-running UNIX philosophy debate illustrates this. Arguments about whether tools should "do one thing and do it well" or handle structured data have traded the same high-level rebuttals for decades. Examining the details — what actually happens when these ideas meet the real world — immediately shows that one set of ideas cannot work as espoused.

Why Copying Fails

Telling someone to write like you fails for two reasons. First, your style may not suit them. Second, and more fundamentally, most people don't understand what drives their own writing — so they can't explain it even when asked. This parallels teaching in other fields. The median dance instructor's verbal instructions don't match what they actually do; students who follow instructions rather than observing and copying end up doing the movement wrong.

Even when people do attempt to copy a successful style, they often copy the wrong things. Natalie Wynn observes that people imitating her video essays add elaborate visuals with no connection to the topic. Her own visuals are conceptually tied to content — Eve in Eden for a video about shame, because Eve was the first to experience shame. The surface detail without the underlying logic is meaningless.

Rob Leatham makes the same point about shooting:

"What keeps me busy in my classes is trying to help my students learn how to think. They say, 'Rob holds his hands like this...,' and they don't know that the reason I hold my hands like this is not to make myself look that way."

Brian Enos describes adopting a champion shooter's system blindly for a year before realizing it didn't fit his own structure and attitude. His experimentation eventually pulled from multiple styles — Isosceles, Modified Weaver, Bullseye — and from shooters and martial artists alike. "Great ideas will arise from a feeling within yourself," he writes. "This intuitive awareness will allow you to accept anything that works for you and discard anything that doesn't."

The same pattern shows up in dance instruction. Students ask whether the left foot crosses in front or behind the right, missing the point: foot placement depends entirely on how the center of gravity is moving, which varies by circumstance. When people don't deeply understand what they're copying, they imitate superficial aspects and miss the fundamentals driving them. No simple rule — "omit needless words" included — can substitute for understanding.

Goals Should Drive Style

Style should emerge from goals, and goals are personal. Possible goals for technical writing include:

  • Explaining a technical topic at a level accessible to almost any professional programmer — branch prediction, malloc, cache partitioning
  • Making the case for a minority opinion: files are difficult to use, public tech companies can pay very well, monorepos aren't stupid
  • Measuring something
  • Discussing interesting phenomena: discontinuities, knowledge transfer difficulties, normalization of deviance

When those goals combine with a preference for detailed discussion, the result is a style distinct from any of the successful bloggers mentioned above.

Process matters as much as style, and a wide variety of processes can work. The author's initial process goals included keeping upfront investment low, improving writing technique over time, publishing only when there was something worth publishing, writing a blog he'd want to subscribe to, and writing on his own platform.

Low upfront investment made sense given that most blogs die quickly — either with a single "I'm starting a blog" post or after a short period of activity followed by dormancy. He used a default blogging software installation before bothering to configure his own domain, which was wise in principle. The specific choice of Octopress was a mistake: annoying setup, breakage, no upstream updates, and absurdly slow builds. Pure HTML or WordPress would have been better.

To improve his writing, he hired a professional editor — whose writing he respected — with instructions to focus on clunkiness and awkwardness, not spelling and grammar. The detailed feedback on each post helped him systematically eliminate recurring problems.

What to Publish and Where

Publishing only when there's something worth saying is a process goal rather than an outcome goal like "one post per week." Outcome goals can be counterproductive for improving quality, since hitting a frequency target on limited time discourages reflection on how to write better.

Things he avoids in his own blog include splitting one topic into many small posts, clickbait titles, and repeatedly covering the same ground without new insight. Writing on his own platform has value because platforms die. Posterous was trendy until Twitter acquired and killed it, breaking links to top Hacker News posts. Blogspot degraded under Google. The trendy platforms when he started — Svbtle and Medium — fared no better. Platform economics eventually compromise the reader experience.

Comments on platforms like Hacker News have short half-lives; a post gets its views in days. A blog post accumulates traffic over years, with most views coming later. Writing in one's own space creates value for both author and readers, rather than primarily for the platform.

No universal advice follows from any of this. The author explicitly avoids telling anyone how to write well. Whether to blog at all is a separate question — the potential value is high and may show up unexpectedly. But how you write should come from your goals and aptitudes.

Refinements Over Time

Several changes to the author's approach emerged over the years. Early on, he'd wait to combine multiple ideas into one post, fearing a single idea wasn't enough. In retrospect, a 2016 post on compensation contained two separately interesting ideas — that compensation might become bimodal, and that programmers are extraordinarily well paid relative to barriers to entry. Splitting them would have reached more readers for both.

He also used to avoid writing things that seemed too obvious. He's since found that many influential posts are exactly about things that are obvious enough to be skipped. More people have said his post on asking questions when you don't know something changed how they operate than any other post — and that topic initially seemed too obvious to write down.

A more recent change is the deliberate use of more examples, prompted by reader feedback. Providing "a lot" of examples reduces the chance readers take away something different than intended. A post about reaching 95th-percentile performance offered only two examples, and many readers filled in the blanks incorrectly — assuming performance far above the 99.9th percentile was merely 95th-percentile. Jamie Brandon made the same change, explaining that concrete examples "pin ideas down so they can be examined properly."

He also works harder on narrative grammar — the continuity that makes writing easier to follow, both at sentence level and in larger structure. This was driven by working with an editor who spots structural issues. Not every reader will follow even well-structured writing: many people answer only half of a two-part question on forums. But writers need not accommodate those who won't read carefully.

Getting Feedback

Feedback can accelerate improvement, but most feedback is counterproductive. Most commenters offer variants of "write it the way I would have written it," which fails for the same reasons general writing advice fails. The feedback tends to be purely superficial — turning idiosyncratic prose into generic Strunk-and-White-ified writing — while ignoring genuine structural problems. Conversely, the common response to substantive feedback is "nah, it's fine," the mirror image of "write it my way." Most people don't know enough about writing to distinguish good from bad feedback.

Hiring an editor whose judgment you trust sidesteps this problem. The author also deliberately takes criticism seriously, even at the cost of spending too much time on clearly bad feedback. As readership grows, the percentage of good comments drops. One common failure mode among established writers is concluding that all critics are bozos. He prefers to ignore feedback from people who regularly give bad feedback — checking what they say about others' writing makes it easy to spot patterns.

Where Ideas Come From

Ideas are everywhere. The world contains enough interesting material that you could write forever. Surf skiing recently produced roughly twenty potential posts after a few weeks — on paddle design, cultural factors affecting adoption, and hull design. This isn't because kayaking is particularly interesting to the author. Everything seems that rich if you look closely enough. Baseball decision-making could have become a long series but never progressed beyond the first post — there's too much other material and not enough time.

What not to do for popularity: If you want a broad audience, certain tactics seem to work even if they aren't necessary. Clickbait titles, especially with swearing or emotionally loaded framing. Self-promotion of prestige. Authoritative tone. Anger-inducing content. High posting frequency. Endorsements. Coverage of hot topics. Outrageous framing. All of these can grow an audience, but none is required.