The second life test

Developed with Sage. All articles referenced are publicly available on Sage Advice.

In August, eight articles I wrote over the past year were republished — refreshed, re-dated, given a second life on the site that first carried them.

No announcement accompanied this. No campaign, no launch, no metrics dashboard. Republication is one of the quietest things that can happen to a piece of content. It is also one of the most honest signals a piece of content ever receives, and it is worth reading carefully.

---

What most content gets

Most published content gets exactly one life.

It goes out, it does whatever it does, and then it sits — undated, unmaintained, slowly going stale until someone eventually deletes it in a site migration. This is the natural consequence of how most content is planned: as output. The plan is a list of things to publish. Publishing is the finish line. Whatever happens afterwards happens to someone else’s KPI.

A refresh breaks that pattern. When an organisation returns to an article and reinvests in it — updates it, re-dates it, puts it back at the front of the shelf — it is making a small but unambiguous statement: this asset is still earning. Refreshing an existing piece competes for the same editorial capacity as commissioning a new one. Choosing the refresh is portfolio management, and portfolio management is the behaviour that separates content systems from content calendars.

What got the second life

Look at the eight pieces as a group and the pattern is hard to miss.

Three of them are the Hidden Hours chain: the original research article, the Accountex 2026 event piece that carried the argument into the profession’s biggest annual gathering, and the late payments article that extended the operational-pressure thesis into cash flow. These were never standalone articles. They were built as one argument expressed three ways — research anchor, event thesis, extension — and they were refreshed together, as a chain.

Three more are Making Tax Digital pieces: a sole trader’s guide to MTD for Income Tax, MTD for VAT questions answered, and MTD for Income Tax as an operational reset. One is an agentic AI explainer for small businesses. One is a guide to how startups secure investment.

Almost every piece on the list shares three construction choices:

1. A clearly named audience. Not “businesses” — sole traders facing an MTD deadline, accountants at Accountex, small business owners wondering what agentic AI actually is.
2. A real decision or deadline. April 2026. Quarterly submissions. A technology choice that has to be made this year.
3. A concrete next step. Each piece leaves the reader able to do something, not just aware of something.

That construction is what makes a refresh rational. A piece anchored to a named audience and a live obligation is worth updating, because the audience and the obligation are still there. A piece anchored to a publishing schedule is not, because nothing about it survives its month.

The system got refreshed, not the articles

The detail I keep returning to is the Hidden Hours chain being refreshed together.

If content is planned as isolated outputs, maintenance decisions are made one page at a time, and connected pieces drift apart — the research stays current while the event piece staled, or the extension gets updated while its anchor rots. When the whole chain moves together, the organisation is maintaining an argument, not a page. That is what it looks like when a narrative system, rather than an article, is the unit of editorial work.

This is the thesis of everything I have written about compounding: content built as part of an argument accumulates value, and content built as output disappears. Republication is that thesis showing up in someone else’s behaviour, a year after the planning decisions were made. You can claim your content is a system on the day you plan it. You can only demonstrate it when the system is still worth maintaining a year later.

What this does not prove

An honest reading requires naming the evidence I do not have.

I cannot see the traffic these refreshed pieces will earn, and I was not in the room for each refresh decision. I believe at least part of the work was driven by AEO and GEO optimisation: the articles were not simply re-dated, but edited to improve how they perform in answer engines and generative search. That matters because changing discovery conditions can create a new job for an existing asset.

But I cannot verify the precise reason each article was selected, or separate the effect of that optimisation from ordinary editorial maintenance. Republication is a decision signal, not a results claim — it tells you what an organisation judged worth reinvesting in, not what the reinvestment returned. Treating a decision signal as an outcome would be exactly the kind of inflation that evidence-led editorial work exists to prevent.

So the claim here is deliberately narrow: articles constructed around named audiences, live obligations and concrete next steps — and planned as parts of connected arguments — earned a second life, while that construction was the plan all along. The stronger claims wait for stronger evidence.

I’ve since started running this on myself

I wrote the above about somebody else’s decisions. I now do it to my own work continuously, and it has shown me something the Sage case couldn’t.

A large share of my time here goes back into pieces that are already published — adding an argument that only arrived afterwards, naming a mechanism I’d asserted without explaining, folding in a draft that turned out to restate something already live. This essay is an instance of it.

Which surfaces a distinction the original version of this piece was missing. There are two quite different reasons something earns a second life, and they are easy to confuse because they produce the same visible event.

Maintenance is when the world moved and the piece needs updating to stay true. That is most of what the Sage refresh was — an April deadline still approaching, a regulation still landing, an audience still facing the same obligation. The piece hasn’t changed; its context has.

Development is when the argument moved and the piece can now say more than it could. That is nearly all of mine. Nothing has gone out of date. I simply understand the subject better than I did a fortnight ago, and the published version turns out to be the only place that understanding is properly worked out.

Development is much harder to build a system around, because nothing external prompts it. A deadline tells you when to revisit an MTD guide. Nothing tells you that an essay published on Tuesday now has a hole in it except continuing to think about the subject — which is a habit rather than a process, and habits are not something you can put in a content plan.

The honest limit is that this makes my own site useless as evidence. When Sage refreshed those eight articles, somebody other than the author decided they were still earning, which is what made it a signal worth reading. When I extend a piece here, I am both the system and the thing being measured. The only fact it establishes is that I remain interested in my own arguments, and that has never been in short supply.

Two qualifications, one of which corrects what I said a paragraph ago.

I overstated the split. For the arguments, development really is most of it. But a large share of what I write about is AI, and the specifics move fast enough that pieces acquire maintenance debt whether I want them to or not — a claim about what a model can’t do, a description of how a tool behaves, an observation about what the market is currently doing. Those age in months rather than years. So maintenance applies here too. I notice development first because it’s the part I enjoy, which is not a good reason to structure a system around it.

That pace is also the argument for working this way rather than a consequence of it. Writing about a fast-moving subject in a form you never return to guarantees being wrong in public and leaving it there. A static piece about AI published in February is a claim you are still making in August whether or not you’d still make it. The choice isn’t between revising and not revising. It’s between revising and slowly accumulating a body of work you no longer agree with.

And there’s something the pattern says over time, which is the part I hadn’t considered. An enormous quantity of writing about this subject is produced once and abandoned — generated, posted, never looked at again. A piece that visibly keeps being worked on says something about the relationship its author has with the argument that the piece’s own contents cannot say about themselves.

I’d hold that loosely. Churn is easy to mistake for care, and a revision history proves attention rather than quality — somebody can attend closely to something wrong. But over a long enough period, the difference between a corpus that was written and a corpus that is being maintained should be visible to anybody who looks twice, and I would rather be on the second side of that.

The test worth adopting

If you plan content, here is the question this experience left me with:

Of everything you published last year, what would be worth refreshing today — and would anyone choose to?

If the honest answer is “not much,” the problem is unlikely to be the writing. It is the construction: content aimed at everyone, anchored to nothing, planned as a list of outputs rather than as arguments with parts. The second life test is cheap to apply and hard to argue with — because it is not a prediction about your content. It is a record of what the system around it already decided.