Where a Blog Post Goes

On September 26, I published Goodbye to the Hard Parts That Never Mattered. It was a response to another post about what AI is changing in software engineering.

By October 24, GoatCounter had recorded 4,818 visits to it. On September 28 alone, it recorded 3,229.

Those are big numbers for this blog. But the part that surprised me most was the list of places people came from.

TLDR. Hacker News. Labnotes. Brutalist Report. LinkedIn. Google. Kagi. Slack. Microsoft Teams.

A newsletter, a discussion site, curated reading lists, search engines, and work chats. All leading to the same Markdown file on my VPS.

The large sources and the small ones

I queried the SQLite database behind my self-hosted GoatCounter instance. These are some of the referral counts from publication through October 24:

Another 1,155 visits had no recorded referrer. That bucket could include bookmarks, email, private messages, and browsers that strip the information. It does not tell me which.

TLDR and Hacker News account for about two-thirds of the article's counted visits. They explain most of the volume.

But twelve visits from Slack are interesting in a different way.

A referrer cannot tell me what someone said when they shared the link. Maybe it supported an argument. Maybe someone thought it was bullshit. Maybe it landed in a channel full of links that nobody had time to read.

I have no idea. I can see the click that reached my site. Everything around it is missing.

The same is true of the Teams-related traffic. It suggests a route through a workplace tool, without giving me a window into the workplace. I find that little trace more interesting than its contribution to the total.

I also found Nathan Pettit responding to the article on LinkedIn. He agreed with parts of it and pushed back on others. Someone else had saved it to Scholion with annotations. Those tell me something a referral count cannot: someone took the time to formulate a counterargument, and someone else took notes. In the traffic data, either interaction could look like another click, if it appeared at all.

Then there are the tracked email clicks. For several days, I have been watching links with HubSpot tracking parameters arrive without being able to identify the sender. I can recognize the distribution technology. I still do not know which email carried the link or what it said.

A different kind of discovery

In May, I wrote about six numbers from this blog's analytics. The biggest single-day spike I found then was 33 visits to a post about Nix shells, SSH, and Claude Code.

The pattern at the time was specific technical problems getting found through search. Someone hits an error, searches for it, and finds the thing I wrote when I hit the same error. A small post keeps being useful months later.

I still like that pattern. A lot of this blog exists because I would rather write down the fix once than figure it out again six months from now.

This post had a different shape. It was an essay about the work itself, and people reached it through newsletters, reading lists, and links shared with other people. Google accounted for just 31 visits in the referral data.

That is an interesting addition to what I saw in May. The technical posts were being discovered through a problem someone already had. This one could turn up among the things someone else thought were worth reading.

The countries were varied too. The United States supplied the largest share, followed by the United Kingdom, Germany, Canada, and Australia. Those are the top five, not the whole list.

There are several kinds of distance in that data. Geographic distance is the one with country names. The distance between my blog and a work chat I will never see is harder to measure.

What the numbers leave out

These are recorded visits, not 4,818 distinct people. They do not tell me how many people finished the post or agreed with it. October 24 was also still in progress when I ran the queries.

The site recorded 45 helpful-click events associated with the article. That is another small piece of feedback, with plenty left unexplained.

I could spend a lot of time trying to turn all of this into a strategy. Find the right subject. Get into the right newsletter. Reproduce the spike.

What interests me more is how many routes showed up at all.

I wrote a response to something I read. It became something other people could read, search for, collect, or pass along. Some of those routes left a name in a database. Plenty probably did not.

A dozen Slack visits will not dominate a traffic chart. They were still someone following a link to something I wrote.

That is a pretty good reason to keep publishing.