Consistency Is Overrated. Persistence Isn't.

TL;DR: “Post every week” is the standard advice for engineers building a public body of work, and it gets the variable wrong. The schedule isn’t what matters. What matters is that you never fully stop during the long flat stretch where nothing measurable happens, and that you keep working one deep topic instead of scattering. A schedule is a tool for staying persistent, and it’s a good one if you can actually hold it. My own archive is the evidence, and not a flattering version of it: 31 posts between 2014 and 2019, then nothing for almost five years, then a restart. The topic survived the gap. The momentum didn’t. So here’s what I’d do differently. Pick a minimum unit you can hold in your worst month, write a restart rule before you need it, and log a handful of leading indicators you control, not the page views you don’t.

“Post every week” is advice about the wrong variable

Every piece of advice about building a public presence as an engineer lands on the same instruction eventually. Pick a cadence and hit it. Weekly posts, a newsletter every Tuesday, a video every other Friday. Don’t break the chain.

I don’t think that advice is wrong, exactly. It’s aimed at the wrong target. It treats the schedule as the thing that produces results, when the schedule is really a mechanism for something else: making sure you’re still showing up in month fourteen, when nobody has replied to anything in a while and you have a real job pulling at the same hours.

That distinction matters because the two fail differently. If you believe consistency is the point, then missing three weeks feels like failing, and a lot of people respond to failing by quitting. If you believe persistence is the point, missing three weeks is a cost you absorb. You publish the next thing and move on. The weekly schedule is supposed to protect you from quitting, and treated as a pass/fail test it gives you a clean excuse to quit.

What my own archive actually shows

I’d rather make this argument from my own record than from anyone else’s, so here it is, pulled straight from the publish dates on this site as of mid-September 2026.

• 2014: 18 posts, starting in May. Ruby, Rails, server hardening, some opinion pieces.
• 2015: 5 posts. One of them was literally about making time to learn while holding down a full-time engineering job.
• 2016–2018: 7 posts across three years.
• 2019: 1 post, in June.
• June 2019 to May 2024: zero. Just under five years.
• 2024: 2 posts.
• 2025: 6 posts, all from late October onward.
• 2026 through mid-September: more than 100 published, with more scheduled.

That’s not a consistency story. That’s a burst, a long taper, a gap long enough that the site was effectively dormant, and then a second burst much bigger than the first. If “post every week” were the real predictor, this archive would be a case study in how not to do it. For the 2019–2024 stretch, it is.

But look at what ran through the gap. My first Phoenix post was a getting-started walkthrough in January 2016. A few months later I posted the slides from an intro to Elixir and Phoenix talk I gave at a local Ruby meetup. Four more Elixir and Phoenix posts followed through 2019, and by then Elixir was my day job: I joined InsideTrack in 2018 and led the Ruby-to-Elixir migration that took it from ten services to six. Then the posts stopped, but Elixir didn’t leave my working life. From 2022 into 2023 I was a senior Elixir engineer at AAMP Global. When the restart turned into real volume in 2026, Elixir came back as one of its main threads. It accounts for more than two dozen of this year’s posts.

So I’ll be precise about what the record shows and what it doesn’t. It shows that a deep topic survives a long publishing gap. What doesn’t survive is momentum. Those five years of silence cost me every compounding effect I’d have gotten from an archive that kept growing. And it doesn’t show that the 2026 burst has paid off, because it’s too early to know. I’m in a flat stretch right now, the same as anyone reading this who started last year.

Is posting on a fixed schedule actually necessary?

No. A fixed schedule is useful if you can hold it for years, and harmful if you pick one you can only hold for a season. What you actually need is a floor: a minimum unit of output you would still manage in your worst month, plus a rule for coming back when you miss even that.

The distinction between “never miss” and “never quit” shows up in the research on habits. The study most people are half-remembering when they say it takes 66 days to form a habit is Lally and colleagues’ 2010 paper in the European Journal of Social Psychology. Volunteers picked a daily eating, drinking, or exercise behavior and tracked it for twelve weeks. The time for participants to reach 95% of their own automaticity plateau ranged from 18 to 254 days. Participants who did the behavior more consistently fit the habit model better, so repetition clearly counts. The part that gets quoted less is the finding that missing a single opportunity to do the behavior did not materially affect the habit forming.

I’d be careful stretching that. Those were simple daily behaviors tied to a cue like “after breakfast,” not a 2,000-word technical essay, and nobody in that study was trying to get noticed by strangers. But the shape carries over. Repetition in a consistent context builds the behavior. An occasional miss doesn’t reset it. What resets it is treating the miss as the end.

That’s where the “don’t break the chain” framing does real damage. It turns a recoverable miss into a psychological failure. A calendar with one red X on it is fine. A calendar you stopped looking at in March because it had three red X’s on it is the actual problem.

So when someone asks me what cadence to pick, my answer is to pick two numbers instead of one:

1. The target. What you’d like to do in a normal month. Weekly, biweekly, whatever you want.
2. The floor. What you will still do in the month your kid is sick, the launch slips, and the on-call rotation lands on you. For most people with a demanding engineering job that’s one short piece a month. Maybe less.

The target is aspirational. The floor is the commitment. The floor is what keeps the gap from being five years.

The flat stretch is where people actually quit

Here’s the part the cadence advice skips. When you start publishing, the results you can see are close to nothing, and they stay there for a long time. You write something you’re proud of and it gets a handful of visits, most of them probably bots. You do it again. Same thing. The effort is real, and the response is low enough that it’s hard to tell it apart from zero.

Nobody can honestly tell you how long that lasts, and I won’t pretend to. It depends on the topic, how much people search for it, how much competition there is, and plenty of luck. What I’m confident about is the structure. Visibility compounds unevenly. Early numbers understate what the same body of work does later. My read is that search engines, AI assistants, and people all need repeated encounters with your name before any of them treat it as a source. That means the flat stretch is exactly when your lagging numbers are the most misleading. It’s also when you’re most likely to quit.

I recognize the texture of this from a different part of my career. I’ve written about spending years on work nobody wanted, like security programs, migrations, and evidence collection, where the visible output of a quarter was a checklist. The payoff was real, but it came late, and for a long time nothing on the outside said it was coming. Public writing has the same cost structure. If you can only keep going while the numbers go up, you’ll stop in the flat part, which is the part that decides everything.

AI made bursts cheap, which makes the flat stretch harder

In 2026 there’s a new way to fail. Drafting is cheap now. I’ve written about committing 4,154 times in six months with AI agents, and the same shift applies to prose. You can produce a month of posts in a weekend, and I’ve written up where AI belongs in a one-person content operation and where it doesn’t.

That makes the burst the easy part. It doesn’t make the flat stretch any shorter. A pile of generated posts landing in the same two weeks, followed by silence, is the old burst-and-quit pattern at ten times the speed. My own July 2026 had 36 posts in it before I pulled back to two or three a week, so this is the failure I’m watching for in my own archive. Volume without validation also burns through your ideas faster than you replace them. Once you’ve published the obvious thirty things about your topic, the thirty-first has to come from work you’ve actually done. The engineers who stand out over the next few years won’t be the ones who can generate the most. They’ll be the ones who are still publishing checked, first-hand work two years from now.

One deep topic beats a consistent scatter

The second half of the argument is about what you’re persistent at. A perfectly consistent schedule across ten unrelated subjects builds less than an irregular schedule on one. That’s the same reason I argue for a narrow front door with the rest of your work visible behind it: one topic gets you remembered, and the range gets you the second engagement.

That’s because recognition comes from repetition in the reader’s head, not on your calendar. Someone who finds your post on Postgres indexing, then a month later hits your conference talk on the same subject, then sees your name in a thread about query planning, now has a reason to remember you. Someone who finds your Postgres post, then a Kubernetes post, then a hot take on hiring, has met three different strangers who happen to share a byline.

The practical version for an engineer is to take one topic you actually work in and push it through several formats:

• The walkthrough post. The thing you figured out, written as the notes you wish you’d found.
• The talk. The same material for a meetup or an internal lunch-and-learn. Presenting it exposes the parts you don’t understand yet.
• The opinionated follow-up. Once you’ve done it in production, what you’d do differently.
• The reference. A checklist, a snippet, or a decision table someone can bookmark.
• The answer. The reply in a forum, an issue thread, or a colleague’s DM that you then turn into a post.

My Phoenix walkthrough in 2016 was followed by a meetup talk on the same subject within a few months, and that was an accident, not a plan. The accidental version still produced the one thread that survived a five-year gap. Done on purpose, one topic across five formats gives you five chances to be encountered and one coherent reason to be remembered.

It also solves the idea problem. You never run out of things to say about a topic you work in every day, because the work keeps producing new material. You run out of things to say about topics you picked because they seemed popular.

What should you measure before anyone is noticing you?

Measure the inputs you control and the early signals that show up before traffic does, and stop checking page views, followers, and rankings month to month. Those lagging numbers are real, but during the flat stretch they’re mostly noise. Worse, they’re discouraging noise that gets people to quit right before things turn. Persistence is necessary, not sufficient, though. When I graded my own blog, the posts were consistent and the distribution didn’t exist, so check the lagging numbers quarterly for whether anyone can find the work at all.

A leading indicator is something that moves before the outcome you care about and that you can influence directly. A lagging indicator is the outcome itself, and it shows up late. If you want to be the engineer people find and trust on a topic, the lagging numbers are traffic, followers, and inbound opportunities. The leading ones come in two groups.

Inputs you fully control:

• Pieces shipped against your floor. Not your target, your floor. Did you clear the minimum this month, yes or no?
• Restart latency. When you missed, how many days passed before you published again? This is the persistence metric. A shrinking number means the habit is holding.
• Depth in your core topic. How many pieces this month were on your one topic, and how many formats did they cover?
• Ideas banked. How many real questions from your work, your team, or your inbox went into the idea list? If this number hits zero, the schedule starts depending on inspiration, and inspiration fails in exactly the months your floor is for.

Early signals you can’t control but can count:

• Replies with a real question in them. Not likes, but emails, comments, or DMs asking something specific. Those people read the whole thing.
• Signups to anything you own. A newsletter or an RSS subscriber count, if you can see it. (I can’t yet. That’s on my own list to fix.) Something that doesn’t depend on a platform choosing to show you.
• Conversations that name a piece. Anyone who says “I read your post on X” in an interview, a sales call, or a Slack thread. Ask how they found it and write the answer down.
• AI assistant citations. Once a month, ask two assistants the main question your topic answers and note whether your work shows up. It’s crude, but it’s a real signal about whether you’ve become a source.

Log it in a plain file, once a month, in five minutes. Something like this:

2026-10
floor met:        yes (2 of 1)
restart latency:  n/a (no miss)
core-topic posts: 2 (post, talk)
ideas banked:     6
real replies:     1
signups:          +3
named in convo:   1 (interview)
AI citations:     0 of 2 assistants

The numbers above are a template, not my results. The point is that every row either goes up because of something you did or tells you something true about whether it’s landing. None of them drops because an algorithm changed on a Tuesday.

After six months of that log you’ll have something page views can’t give you, which is evidence of whether you’re persisting and whether anyone is starting to notice. That lets you judge the flat stretch on data instead of mood.

Write the restart rule before you need it

The last piece is the one I wish I’d had in 2019. Every long gap in my archive started as a short one. Nobody decides to stop writing for five years. You miss a month because of a job change, then the next piece feels like it has to be a big comeback to justify the silence, and the bar for “good enough to break the silence” keeps rising until it’s out of reach.

The fix is a one-line rule, written down now, while you’re motivated:

If I miss my floor, the next thing I publish is the smallest useful thing I know, within seven days, with no mention of the gap.

No apology post. No “I’m back” announcement. No saving it for something worthy. A short note about a bug you fixed that week is enough. The rule exists to make restarting smaller than the gap wants it to be.

This connects back to why persistence beats consistency. A consistency mindset makes the restart feel expensive, because you have to own the broken streak. A persistence mindset makes it cheap, because the only question is whether you’re still going. You are. Publish the small thing.

What to do this month

1. Set a target and a floor. Write both down. Only the floor is a commitment.
2. Name your one topic. The one you work in every day, so the work keeps producing material.
3. Plan one piece in a second format. Turn your best post into a talk, a checklist, or a long answer somewhere your peers ask questions.
4. Write your restart rule. One sentence. Put it where you’ll see it when you miss.
5. Start the monthly log. Inputs and early signals only. Check the lagging numbers quarterly at most.

My own archive says a topic survives a long silence and momentum doesn’t. I’d rather you keep both.

Read this next: The Career I Built on Work Nobody Wanted is the same trade in a different arena. It covers years of low-visibility work that paid off late, and the honest cost of taking that bet.