Rerun The Winners

You've probably published more good content than you realize.

Two years of newsletters. A few hundred social posts. Some blog articles you were proud of at the time. Maybe a podcast or a YouTube channel that ran for a while. It's all sitting there, and most of it hasn't been seen by anyone since the week it went out.

Meanwhile you're sitting down on Monday morning trying to think of something new to say.

That's backwards. The best performing piece of content you'll publish this month is probably something you already wrote. It just needs to be found, updated and put back in front of people, most of whom never saw it the first time.

Here's the part that changed last week and makes this worth doing right now. On September 22, OpenAI released GPT-6 Luna at ten cents per million input tokens. A million tokens is roughly 750,000 words. That's somewhere around three hundred full length newsletters for a dime.

Processing your entire archive, reading every piece, scoring it and tagging it, used to be a real expense. Now it costs less than the coffee you're drinking while you read this.

This sounds like a lazy person's strategy. It isn't. There are three reasons your archive is worth more than your blank page.

Most of your audience has never seen it. If your list has grown at all in the last year, a big chunk of your current readers joined after your best pieces went out. For them, your old work isn't a rerun. It's new.

It's already been tested. You know which pieces got opened, clicked, replied to and shared. A new piece is a guess. An old winner is data.

Your thinking has gotten better since you wrote it. Going back to a piece from eighteen months ago and updating it with what you know now almost always produces something sharper than either the original or a brand new piece would have been.

The catch has always been the effort of finding the good stuff. Nobody wants to reread three hundred issues to figure out which twenty are worth revisiting. That's the part AI now handles for almost nothing.

Before you can score your archive, you need it in a format a model can read.

If you publish on beehiiv, you can export your posts. Most platforms have some version of this. Pull the title, the publish date, the full text, and whatever performance numbers you have: opens, clicks, replies. Put it all in a spreadsheet, one row per piece.

For social posts, most schedulers let you export history with engagement numbers. For a podcast or video channel, you want transcripts, and if you don't have them already, transcription is cheap now too.

Don't overthink the format. A spreadsheet with a column for the text and columns for the numbers is plenty.

This step takes an hour or two. It's the only tedious part of the whole system.

Now you're going to have a cheap model read every single piece and score it. Not on quality, because a model's opinion of your writing quality isn't very useful. You're scoring on specific, checkable things.

Here's the prompt I use. Run it once per piece, which is easy to automate in Make.com by looping over the rows in your sheet.

Paste the piece underneath. Store the output back in the same row.

At current pricing on a model like Luna, running this across three hundred pieces costs well under a dollar. Even on a much more expensive model, you're talking about a few dollars. There's no cost argument against doing this anymore.

Now sort. You're looking for the overlap between two things: pieces that performed well when they went out, and pieces that scored 4 or 5 on evergreen.

High performance plus high evergreen is your rerun list. These are proven ideas that still hold up. Most people end up with somewhere between ten and thirty of these in a decent sized archive.

High performance plus low evergreen is interesting too. These were popular because they caught a moment. The topic might deserve a fresh take with today's version of the news.

Low performance plus high evergreen is your hidden list. Good ideas that didn't land, maybe because of a weak headline, bad timing, or a smaller list at the time. These are worth a rewrite with a new angle and a better headline.

Low performance plus low evergreen, leave alone.

FROM THE AI NEWSROOM

The exact systems behind everything in this issue. The audit sheets, the routing logic, the templates and the review cadences, built out step by step so you can copy them straight into your own stack. One time, forty seven dollars.

Get the Blueprint for $47

.....

This is where most people get lazy and it's where the value actually comes from.

Don't just resend the old piece. Update it. The "dated_references" field from step two is your to do list. Every product name that changed, every price that moved, every statistic that's now old, every "last week" that's now two years ago.

Then go further. Ask yourself three questions about each piece:

What do I know now that I didn't know when I wrote this? There's almost always something. A mistake you made since, a better example, a result from a client, a nuance you missed.

What's happened in the world that makes this more relevant, or less? Sometimes the news cycle has made an old piece suddenly urgent.

Is the headline as good as it could be? Your headline writing has probably improved. Rewrite it.

Updating a piece this way takes thirty to forty five minutes. Writing a new one from scratch takes two or three hours. And the updated piece often performs better because it's built on an idea you already know works.

A quick note on honesty with your readers. If you're sending something that ran before, you don't need to hide it. A line like "I first wrote about this in 2025 and I've changed my mind on two things since" is a hook in its own right. People like watching someone's thinking evolve.

One good long piece contains five to ten smaller pieces. Your reusable_assets column from the scoring step has already found most of them.

A checklist from a newsletter becomes a carousel. A framework becomes a short thread. A single strong sentence becomes a standalone post. A story from the middle becomes a video script.

Here's the prompt for that step:

Load the results into Buffer and spread them out over a few weeks. One updated piece can fill a week and a half of social without you writing a single new idea.

Here's how this fits into a real week without taking over your life.

Pick one day a month, a couple of hours, for the archive work. Score anything new from the past month, then pull the next three pieces from your rerun list and update them.

Then mix. I'd run roughly one updated piece for every three new ones. That ratio keeps your content fresh for long term readers while giving new subscribers access to your best proven material. If you're going through a busy stretch, go heavier on the reruns. That's what they're for.

Keep a simple log of what you reran and when, so you don't send the same piece twice in six months. A column in your archive sheet with "last rerun date" is all you need.

Be realistic about the scoring. It's useful, but it's not magic.

The evergreen score is a guess. Models tend to overrate generic advice as evergreen and underrate pieces with specific examples, when in practice the specific examples are usually what made the piece good. Use the score to sort, then use your own judgment on the top of the list.

The dated references list will miss things. It'll catch product names and statistics reliably, but it may not know that a tool you recommended got acquired, or that your own opinion changed. Always reread before you rerun.

And the model can't tell you why a piece performed. It can't see that it went out on a holiday weekend, or that someone big shared it, or that the headline was the whole reason people opened it. You have to bring that context yourself.

There's a bigger reason to build this system now.

AI search is increasingly how people find content. When someone asks ChatGPT or Perplexity a question in your area, those systems look for clear, current, specific answers. An archive full of pieces with stale product names and old statistics looks less current than it is. Updating your best pieces keeps them in the running.

And the volume of new content being published is exploding, because everyone else has AI too. In a world where new content is cheap, content with a track record is more valuable, not less. Your archive is a moat nobody else can copy, because it's yours.

Most people treat content like it expires the week it's published. The ones who treat it as an asset they own, something to maintain and reuse, will build more with less effort than everyone chasing the next new idea on Monday morning.

FROM THE AI NEWSROOM

For operators who would rather build it with someone than build it alone. The full stack, installed alongside you, with the decisions made in the room instead of guessed at. Ninety seven dollars.

Join the Accelerator for $97

.....

Export your last twelve months of published content into a spreadsheet. Title, date, text, and whatever numbers you have.

Run the scoring prompt on the top twenty performers only. That's a manageable first batch and it'll cost you pennies.

Pick the single piece that scored highest on evergreen and did well when it went out. Update it this week, rewrite the headline, and send it.

Then watch what happens compared to your average new piece. That comparison will tell you exactly how much of your archive work to do from here on.

Jordan

The AI Newsroom | Practical AI for people with a business to run.