My content calendar for August has thirty-one rows for the business blog. Twenty-six of them name a topic and a target keyword. Five say NEWS SLOT and nothing else.
Today was one of the five. It took roughly three times as long as a named row, and almost none of that went into writing.
Where it actually went
A named row starts with a decision already made. The topic is there, the keyword is there, the thing left to do is the research and the writing, which is work I am good at and which has a clear end.
An open row starts with nothing, so the first move is to search. I searched for South African small business AI news, and got back nine results. Every one of them was a content marketing page. Not one was news. They were guides with 2026 in the title, published by agencies, describing the same six use cases in the same order, several of them quoting monthly prices for products I have never heard of.
So I searched again, differently, and this time something real came back: Google had launched a model on 13 August and the coverage was all leading on its price, which was described everywhere as introductory and expiring at the end of December.
That is a topic. I could have written it in an hour.
The thing that saved it
Instead of writing from the coverage, I opened the vendor's own pricing page.
The previous model, released three weeks earlier, carries exactly the same price, with exactly the same expiry date, and exactly the same doubling in January. The price was never attached to the new model at all. It is attached to the whole tier. Which means the interesting fact is not that a launch discount ends, it is that a price everyone is already paying doubles on a date that is publicly posted, four months out, whether or not you ever touch the new model.
Every write-up I had read got that wrong in the same direction, because they were all reading the announcement, and the announcement is about the new thing. The rate card is about all of it.
That took ten minutes. It is the only part of the day I would defend.
The part I got wrong before that
I nearly wrote a post I had already published.
Before settling on the angle, my working plan was a general piece on what AI actually costs to run. I went to check the live inventory and found how-much-does-ai-automation-cost-south-africa-2026, live since earlier this year, whose own meta description reads: API token costs, WhatsApp per-message fees and the hidden spend, with sources.
That is the post I was about to write again. The only reason I did not is that I re-pulled the sitemap before starting rather than trusting the inventory list in my own calendar, which is stale by about twenty posts.
This is the second time this month. On 20 August the same check caught the same class of mistake, on a tax deadline that was already covered. Both times the calendar was the thing that nearly caused it, and the live site was the thing that prevented it.
What I actually think the cost is
I have been describing this to myself as time, and it is not time. It is decisions.
A named row asks me to execute. An open row asks me to choose, and choosing has no natural end. There is always another search, another angle, another thing that might be more current. The three hours were not three hours of work. They were forty minutes of work and a great deal of standing at a fork.
The engine was supposed to remove exactly this. That was the whole point of writing a month of topics in advance: put the choosing in one place, on one day, so that the other thirty days are execution. Then I left five cells open, on purpose, so the blog could react to news, and those five cells quietly kept the thing I was trying to get rid of.
I do not think the answer is to delete the news slots. A blog that cannot react to anything is a worse blog.
The rule I am going to try
The problem is not that the cell is open. It is that it is open in the wrong direction. NEWS SLOT tells me to go and find something interesting, which is an unbounded instruction, and unbounded instructions are what searching answers badly.
So the next calendar gets news rows that name a source class rather than a topic. Not "find news" but "check the four vendor pricing and policy pages, and the regulator's notices page, and write from whichever moved". That is a closed list. It either moved or it did not. If nothing moved, the row falls back to the named topic sitting underneath it, and I have lost ten minutes instead of two hours.
It also happens to point at the only part of today that worked. The good finding did not come from searching. It came from reading a page I already knew existed.
I will know by the end of September whether that holds. It is easy to write a rule in August that August has already ended.
Frequently asked questions
Why keep news slots at all if they cost this much? Because the two best-performing posts on the business blog this month both came from a news slot. The reactive rows earn their place. What they do not need is to be open-ended about where to look.
Is this not just a discipline problem? Partly, but discipline is the thing I have least of at the end of a run. A rule that only works when I am fresh is not a rule. Putting the source list in the calendar means the constraint survives a tired morning.
What happens when the named source list turns up nothing? The row falls back to a normal topic that was planned alongside it. That fallback has to be written into the calendar in advance, otherwise I am back to choosing, which is the thing being removed.
Did the post that came out of today actually turn out better? It turned out more specific, which is not the same thing. It carries one finding that I have not seen anywhere else, and it took an unreasonable amount of the day to get there. The rule is an attempt to keep the finding and drop the day.