I Graded My Blog Against a 2014 Marketing Course

TL;DR: I own a copy of John Sonmez’s 2014 package How to Market Yourself as a Software Developer, and this month I read all of it and graded sublimecoding.com against its core claims. The technical side of my site passes easily, and my publishing cadence meets what the course asked for. I failed the two things it cared about most: I have no email list, and I barely promote what I publish. The numbers are ugly. 63 search clicks in 92 days, 126 of 145 pages with zero clicks, and brand search close to nothing. A lot of the course’s platform advice is dated or wrong now, and it says nothing about AI answer engines, which it couldn’t have. The worst part is that last week I published a ranking that put distribution in S tier, on a site with no owned audience. Below: what held up, what didn’t, and what I’m changing. None of the changes have results yet. I’ll report those in a follow-up.

Why grade a blog against a 12-year-old course?

Because old advice is a good control group. A 2014 course can’t be chasing this year’s algorithm change or selling this year’s AI tool, so whatever still holds up in it is probably structural. And whatever I’m skipping that it treated as basic isn’t a clever modern shortcut. It’s just a gap.

The package is John Sonmez’s developer-marketing bundle from Simple Programmer, announced in March 2014 as a mix of books, video courses, and interviews with well-known developers. It’s still listed on his products page. I own a copy, and this month I went through all of it: six PDFs and roughly nine hours of video. I’m not going to reproduce it or walk through it chapter by chapter. That’s his work, and you can buy it. What I’m doing here is commentary: the core claims in my own words, graded against a site I run and have real numbers for.

The reason I did it now is simple. I’d been publishing two to three posts a week and getting almost nothing back, and I wanted an outside checklist that wasn’t written by me.

The receipts: what sublimecoding.com looks like today

Before grading, the actual state of the site. All of these come from my own Search Console exports and a read-only review of the codebase.

• Search traffic: 63 clicks and 5,612 impressions over a 92-day window. That’s a 1.12% click-through rate.
• Pages that earn nothing: 126 of 145 pages with impressions got zero clicks in that window. Only 19 pages got even one.
• Brand search: 7 branded queries and 90 impressions in 92 days, and 64 of those impressions were me running site: searches on my own domain. Almost nobody is typing my name into Google.
• Email capture: none. No newsletter, no signup form, no list provider anywhere in the code. The only way to reach me is the contact form or a mailto: link, and both are one-off messages. Neither builds a list.
• Promotion: my LinkedIn promo log has three posts since July. Over the same stretch I published about 70 posts, 36 of them in July alone.
• Cadence: 2-3 posts a week since I pulled back from near-daily in August.
• Technical SEO: strong. Structured data, FAQ blocks, an llms.txt, five topic pillars, per-post title and description limits enforced by tests that fail the build.

A more recent 28-day export showed 71 clicks on 4,241 impressions. So it’s moving, but from a very small base, and I’m not going to call that a trend yet.

The site is also well built on the writing side. Posts go through an agent-assisted pipeline: demand validation before drafting, a drafting step, then an editorial gate that checks for invented facts and uncited claims, and a mechanical preflight. Which model runs which step lives in one routing file. Writing a good post is no longer my bottleneck.

That’s the setup for the grade. Production is solid. Distribution barely exists.

Why do 126 of 145 pages get zero clicks?

Mostly because nobody gets a second encounter with the work, and search alone isn’t delivering the first one at any real volume. A page that ranks on page three for a query nobody types gets impressions and no clicks. A page that ranks fine but has a title nobody picks gets impressions and no clicks. And a reader who does click has no list to join and no feed where they’ll see me again, so a good visit leads nowhere.

Some of those 126 are archive posts from the Rails and ColdFusion years, and nobody should expect them to earn clicks. That doesn’t explain the money pages. My /playbook page had 705 impressions at an average position around 9.6 and zero clicks. /resume had 276 impressions around position 8.7 and zero clicks, and /start had 94 around position 9.4, also with zero. Those are pages built to convert, sitting on the first page, and nobody picked them. The vCISO cost calculator is a different problem: it drew over 600 impressions from deep in the results, which says the demand is real and the page isn’t ranking.

The 2014 course’s assumption was that more content brings more search traffic. That held in 2014. On my site, in 2026, the zero-click count is the counterexample. Volume without distribution gave me a big pile of pages that search engines know about and people don’t.

Where the 2014 advice held up

The durable parts of the course aren’t the tactics. They’re the claims about how attention works, and those haven’t changed.

Start the email list before it feels worth it

This is the course’s most load-bearing claim, and it’s the one I failed hardest. In the tutorial on getting traction, Sonmez tells viewers to start a mailing list right away, and says it’s advice he wishes he had followed from the start. The reasoning holds: a list is the one channel where no platform decides whether your reader sees what you sent. Every reader who comes once and leaves without a way to hear from you again is gone.

The usual reasons for skipping one are easy to reach for. The list would be tiny. A newsletter is another thing to maintain. The site is a consulting funnel, not a media business. All of that sounds reasonable. None of it survives the numbers above. If my brand search is near zero after roughly 65 posts since July, then the people who liked a post didn’t come back on their own. A list is the mechanism that brings them back, and I never built one.

The 2026 version is stronger, not weaker. AI answers on the results page mean fewer clicks make it through (more on that below), so each visitor who does show up is worth more to keep.

A post isn’t done until it’s promoted

The course treats promoting every post as basic, and it’s right. Three LinkedIn posts against about 70 published is not a promotion strategy. Most of what I published got whatever search gave it, which, per the numbers above, was very little.

This one hurts more in 2026 than it would have in 2014. Back then, writing the post was the expensive part and promotion was the chore you might skip when tired. My pipeline made writing cheap. Drafting a native LinkedIn version of a post takes minutes. When the expensive step is automated and the cheap step is still skipped, the skipped step is a choice, not a time problem.

People need several encounters before a name sticks

The course keeps returning to repeated exposure: people have to run into you more than once, in more than one place, before they remember you. My brand-search number fits that. A reader who finds one post through search has, on my site today, no second encounter planned. No list, no regular feed presence, no appearances on anybody else’s show or newsletter. So they read one post and forget the name.

Titles have to earn the click

Catchy, honest, tested headlines were part of the course’s blogging advice, and I’ve only done half of it. My titles are optimized for length compliance: a test fails the build if the SERP title runs past 70 characters. None of them are tested for whether anyone clicks. A 1.12% CTR and money pages at position ten with zero clicks say I’ve been solving the wrong constraint.

Traction takes years, and cadence matters

Here I pass. The course asked for at least weekly and suggested three a week to start. I hit near-daily in July and sit at two to three a week now. It also warned that traction takes years, which is the honest counterweight to everything above. My site has been in its current form for a matter of months, not years. Some of the ugly numbers are just early. But persistence doesn’t explain brand search this low after this much output. Consistency was never my problem. Nobody seeing the posts was.

Where it’s dated or wrong now

A lot of the course’s mechanics belong to 2014, and grading me against them would be unfair to both of us.

• Specific platforms. Some of the social advice leans on Google+ communities, and Google shut down consumer Google+ on April 2, 2019. Follow-for-follow-back growth tactics feel like spam to me now. The principle underneath, going where your audience already gathers, holds. The venues moved to LinkedIn, niche newsletters, podcasts, and invite-only chat communities.
• Hosting and tooling. The shared-hosting and WordPress walkthroughs don’t apply to a Phoenix app on Fly.io. The rule underneath, own your domain and your platform, does apply, and I already follow it.
• Comments as a growth channel. Commenting on popular blogs to get noticed isn’t a real channel anymore, and blog comment sections on a site like mine are a moderation cost with no payoff. Showing up where people already gather still works. The comment section just isn’t that place now.
• Backlinks as the main lever. The course’s SEO model was: write great content and earn links. Links still matter. But discovery now also runs through ranked feeds and AI answers, and guest posting for links is mostly spent: Google’s spam policies name large-scale guest posting campaigns with keyword-rich anchor links as a link scheme.
• Podcast scarcity. The course framed developer podcasting as an underused channel. Nobody would call it underused today. Guesting still works, but the advantage now comes from fitting a niche show well, not from the channel being empty.
• “More content, more traffic” in its pure form. Covered above. My own site disproves it.

None of this makes the course bad. It makes it 2014. The claims about attention held up. The advice about specific platforms didn’t.

What it couldn’t have known: answer engines

The biggest gap isn’t anything the course got wrong. It’s a channel that didn’t exist when it was written. For a lot of searches, the answer now shows up on the results page before anyone clicks. Pew Research tracked real browsing and found Google users clicked a traditional result link in 8% of visits when an AI summary appeared, versus 15% when one didn’t, and clicked a link inside the summary in just 1% of visits.

That changes what “discovery” means. For a lot of queries, being named in the answer is the new version of a visit. It’s also why I’ve invested in the parts of the site that make content easy to lift and cite: FAQ blocks, structured data, question-shaped headings, and an llms.txt.

Here’s the honest grade on that investment: incomplete. I can’t yet pull AI crawler hits out of my server logs in a way I trust, so I don’t know whether any of it is getting cited. I built for a channel I can’t measure. It’s the right bet, I think, but it’s still a bet. And it makes the email list more important, because a reader who never clicks through never meets a signup form.

Does 2014 developer marketing advice still work in 2026?

Mostly yes on principles and mostly no on platforms. The ideas about attention held up. The specific channels didn’t. And AI answer engines are a gap no 2014 course could have filled. Here’s my site graded against the claims that mattered:

Claim from the course	Still true?	My grade
Own your domain and platform	Durable	Pass
Publish on a steady schedule	Durable	Pass
Hold a quality bar	Durable	Pass, with an editorial gate
More content means more search traffic	Wrong in pure form	Passed the input, failed the outcome
Promote every post	Durable	Fail
Start an email list on day one	Durable, stronger now	Fail
Repeated exposure builds a name	Durable	Fail, near-zero brand search
Test headlines	Durable	Half: length-tested, not click-tested
Keep one consistent identity	Durable	Unresolved
Traction takes years	Durable	Too early to grade
Discovery through AI answers	Not covered	Built for it, can’t measure it

The pattern is uncomfortable because it’s so clean. I did every part that looks like engineering and skipped every part that looks like marketing. That’s a very predictable failure for an engineer. It’s also the one I’d expect to catch quickly in someone else’s plan.

The distribution gap on my own site

Last week I published a ranking of 13 ways to make money with AI and put media and distribution in S tier, because an owned audience makes every other item on the list cheaper and faster. I still believe that. I just hadn’t built one. Those same checks now run as a free audit skill against any blog: promotion coverage, pages with impressions and no clicks, and where capture exists.

The advice in that post is right. My own site doesn’t follow it. Saying so here is cheaper than letting a reader notice it first.

What I’m changing

These are plans, not results. I’m listing them so the follow-up has something specific to measure against.

An email list, built around one buyer problem

Not a general “subscribe for updates” newsletter. The course’s point was to start one. The 2026 refinement is to make it narrow and tied to a specific problem, so the signup offer matches why the reader showed up. My best candidate is the vCISO cost cluster. The vCISO cost calculator and its related posts, like the vCISO math for AI founders, draw real impressions from people with a real budget question and turn almost none of it into anything. The plan is a signup right after the calculator shows its result, since that’s the moment a reader has just gotten something useful.

It would also close a measurement gap. As of early August I couldn’t attribute a single inbound lead to a single post. A list with source tags would give me a signal long before the contact form does.

Promotion as part of publishing

Each post will get a native version for LinkedIn and one other channel before it counts as published. Written for the platform, not a link dropped with a sentence on top. The same pipeline that drafts posts can draft these. A human still decides what ships.

Borrowing other people’s audiences

A new name doesn’t have an audience to tell, so the early leverage is in other people’s. For each pillar I actually care about (AI engineering, startup security, Elixir), the plan is to find the curated newsletters and niche podcasts whose readers overlap with mine, and pitch them something built for their audience instead of a generic “I’d love to come on.” I haven’t done any guesting yet. I’ll write about it after I do, not before.

One brand

My site is split. The favicon is a “Jared Smith” monogram, the domain is sublimecoding.com, the organization in my structured data is “Sublime Coding,” and there’s an unused Sublimecoding logo in the static assets. On top of that, one of my five pillars is Omarchy on Apple Silicon, which I haven’t explicitly connected to the founder-and-engineering-leader promise. The course was right that your identity should be consistent everywhere, and in 2026 a split one also gives search engines and language models two fuzzy records instead of one clear one. I’ve picked my own name. The site sells my judgment, so the brand should be me, on this domain. The next step is making everything agree with that and connecting Omarchy to the rest. The thread is already there: the Omarchy series is headed for whether it passes SOC 2 on a company laptop, which is the same question I answer for founders.

Titles tested for clicks

The money pages that rank around position ten and get zero clicks will get rewritten titles, with the date noted, so I can compare click rates a month later instead of guessing.

What I’m not claiming

No subscriber count, no traffic lift, no leads. None of those exist yet, and I’m not going to write this post as if they do. The follow-up will report what the list, the promotion, and the title changes actually did, including if the answer is “not much.”

The most useful thing an old course did for me wasn’t new information. It was an outside standard that had nothing to do with my preferences, and it showed me I’d built a very good publishing system for a distribution channel I never set up.

If you’re a technical founder whose product is well built and whose go-to-market plan is “we’ll write good content and they’ll find us,” you probably have the same gap I do. The engineering and security side of that company is where I can help directly, and that’s the work I do. The distribution side, as this post shows, I’m still working out in public.