How the content engine works
It comes up with the article ideas, writes the drafts, takes your edits, and gets a little better every time you correct it. You stay the editor. It does the typing.
A look under the hood: where the ideas come from, how a draft moves through review, and the quiet loop that turns your edits into rules for next time.
The short version
Once a week it proposes a short list of researched article ideas. You accept the ones you want. Each accepted idea becomes a draft, written in your brand voice and checked by an editor before you ever see it. You read the draft in a normal Google Doc, leave comments or edits, and it revises until you accept. When you accept, it publishes and then quietly studies what you changed, so the next draft already knows it.
That's the whole idea. The rest of this page is what each of those steps actually does, and where you get to stay in control.
Four stages move an article forward. The fifth, in green, runs after it ships. That last step is the difference between a writing tool and an engine that compounds.
Where the ideas come from
You don't start with a blank page. Once a week the engine hands you a short list of researched ideas, each with a real keyword behind it, and you pick the ones worth writing.
Every Monday morning it builds a fresh batch. It reads what you've already covered so it never proposes a duplicate, researches roughly a dozen candidate angles across several topic areas, checks real search demand on each, and then cuts hard. What lands on your sheet is the six that survived, not the twelve it started with. Each one arrives with a working title, the primary keyword, supporting keywords, and a one-line reason it's worth doing.
Nothing gets written yet. The batch sits there with the decision column blank, waiting for you. You accept the ideas you like and ignore the rest. That blank column is the first place you're in charge: the engine proposes, you commission.
One column moves every article
There's no project board to learn and no buttons to wire up. A single status column in a Google Sheet steers every article through its whole life, and changing one cell is how you give the next instruction.
When you accept an idea, its row flips to Ordered. The engine picks it up, marks it In progress, and writes. When the draft is ready it sets the row to In review and the work lands back with you. You read it, and you decide: type Accepted and it moves to publish, or type Revise and it goes back for another pass. When it's live, the row reads Done. That's the entire interface. One column, plain words, and you're always the one who sets the next state.
The green path is the one you trigger most: type Revise and the draft loops back for another pass. Every other move is the same idea, one word in one cell.
How a draft gets written, and checked
A draft isn't written in one shot. It's researched first, written in your voice second, and then put through an editor that can send it back before it reaches you. You only see drafts that already passed a bar.
When a row goes In progress, the work runs in order. First a researcher gathers the facts and verifies each one against a live source, so the draft states things that are actually true, not things that sound true. Then a writer produces the article in your brand voice, following any steering you left on the row. Then an editor reads the whole thing as a gate, not a suggestion box. If the voice is off, a fact is stated without backing, or your instructions were ignored, the editor sends it back for a rewrite.
That rewrite happens once. If the draft still doesn't clear the bar on the second try, the engine doesn't keep grinding or quietly ship something weak. It flags the row for a human to look at. The point of the gate isn't to be clever, it's to make sure the thing that reaches your inbox is worth your reading time.
Your edits are the instructions
When the draft lands in review, you treat it like any Google Doc: leave comments, rewrite a sentence, change the angle. The engine reads exactly those edits and revises only what you touched. Nothing else moves.
This is the part that feels different from a one-shot writing tool. You're not re-prompting and hoping. You're editing a document, the way you already work. When you set the row to Revise, the engine reads your inline comments and your notes, looks at the sentences you changed by hand, and makes those specific changes, in the same doc. It doesn't rewrite the parts you left alone. Then it hands it back for another look. You can go around as many times as you want, and every round is saved.
The loop that makes it an engine
Here's the quiet part. Every time you accept an article, the engine studies what you changed along the way and turns the durable lessons into rules. The next draft starts from those rules. It gets better at sounding like you without anyone retraining a model.
When a row hits Done, a final step runs that you never see. It reads the whole journey of that article: the comments you left, the sentences you rewrote, and especially the quiet edits you made right before accepting, which are the most honest signal of all. From that it tries to distill general lessons, not one-off fixes. "This title was too long" isn't a lesson. "Cut hedging adverbs in the opening" is.
It holds those candidates to a strict bar, so most rounds produce nothing, and a good round produces one or two. Anything that survives gets written down as a dated rule with the exact edit that justified it, and that rule steers every future draft. An article you accept without a single change writes nothing at all, because the engine reads that as "you did it right." This is why the tenth article costs you fewer edits than the first.
No model is retrained. The engine just keeps a short, growing list of rules earned from your real edits, and applies them to the next draft. That's the whole trick, and it's why the work compounds.
Most writing tools start from zero every time. This one remembers what you taught it last week, so your editing gets lighter as the engine gets sharper.
The dials: tuning it to your operation
The defaults are deliberate, and they're adjustable. You don't touch code. You tell us where you want to sit, ideally against a real article, and we set them.
| Dial | What it controls | Loosen or change it for |
|---|---|---|
| Voice profile | The master dial. Defines how it writes: tone, structure, how often it calls a reader to action, what it never says. | A different brand, a second site, a sharper house style |
| Ideas per batch | How many researched ideas land on your sheet each week. Default is six, drawn from a wider candidate pool. | A faster publishing cadence, or a slower one |
| Editor strictness | How hard the editor gate pushes back before a draft reaches you. Default is strict, with one rewrite allowed. | Letting solid drafts through faster, or holding an even higher bar |
| Per-article steering | A note on the row that aims a single article: the angle, the audience, the call to action you want this time. | One-off pieces that shouldn't follow the usual pattern |
What you get, and who stays in control
What you get is a steady supply of finished drafts in your own brand voice, each one already researched, edited, and waiting in a Google Doc for your yes. What you keep is the editor's chair. Nothing publishes that you didn't accept.
The engine never decides on its own that something is good enough for the world. It proposes ideas, you commission them; it writes drafts, you approve them; it learns from your edits, and you set the direction those edits point in. The machine does the research, the typing, the fact-checking, and the patient remembering. You do the thing only you can do, which is decide what's worth saying and whether this said it. That division of labour is the point, not a limitation.