A weekly education newsletter for under $1

From a team to a switch

At ContentKing we had a curated newsletter about industry news. It was well written and easy to read, and it became a media channel next to our content marketing. It also helped promote the product. But a newsletter like that used to take a team. People researched the news, curated it, wrote it, designed it on brand, and ran the sending tools, with an email marketing manager over all of it. With AI agents, I thought work of that quality could now be done fully automatically. For Flip Education, that meant a channel you switch on once, and it keeps working.

What this makes possible

I'd say a curated weekly news channel suits almost any brand that wants to stay in the minds of its buyers. It's almost like having a fairly senior employee on call for this exact job, ready whenever you want it, for any project. Which means such a channel is no longer a decision about hiring.

The system has a few parts that fit together simply. Stories come in from the week's news, and a saved copy of every source is kept. AI agents pick the stories and write, every fact is checked, and the result becomes a web page, an email, a short video and a version for each group of readers, with my calls on taste and brand risk. Every output is made from the same checked facts, so a fact is checked once and used everywhere.

The challenge

I was fairly sure the newsletter part would work. The video was the big unknown. My bet was a model released that same week, one that writes the code that draws each frame of the video. A week earlier, I don't think any model could have done this at this quality and cost.

How an issue is made

First, scripts and AI agents collect the week's education news and save a full copy of every page, so each fact traces to a page that can't change. AI agents then score every story against written rules, and an editor agent picks with a written reason, so every pick follows rules that can be checked.

AI agents write each item. Every claim is matched word for word to a saved copy of its source before anything is published. A script does that check, with no AI in it, and a separate AI agent, not the writer, reads each item for meaning. A claim that fails is repaired or cut, so no made-up fact reaches a reader.

Scripts build the web and email editions, code written by AI agents draws the video frame by frame, and an AI voice narrates it, all from the same checked facts. A matching script scores every story for each reader and decides to keep it, have it re-told with figures that concern them, or swap it for a closer story. Every changed story goes back through both checks, because a re-told story is where mistakes get in, so each group gets stories that fit.

People decide only taste and brand risk. That means the voice, the design, and which sensitive stories to run.

A copy shaped to each reader

Some tools already shape each copy to its reader by learning from past clicks. I wanted the copy shaped from what the product already knows, so it works from the first issue. I tested it on that week's saved stories.

The test used three reader groups the product already knows from its sign-up form (a high-school science teacher, an early-grades teacher and a high-school history teacher).

For the science teacher, a bill about young children became a report on state support for high school redesign. Every reader copy passed both checks, and every mistake the meaning check caught was fixed, or the part cut, before the copy passed. The test took under two hours and used no paid services.

What changed

The method is now written down, so any of my projects, or Flip Education's, can reuse it without starting from zero.

Who should run this

I'd say this fits a company whose market has no good curated news source for its people. Is there a strong newsletter in your market that curates the news for its readers, so they can easily follow what matters? If not, that space is open. And ask what your own product already knows about each reader.