The edit is where AI content becomes worth reading
Every session so far - the idea, the repurposing, the long-form, the social, the visuals, the pipeline - hands you a fast draft. None of it earns the reader on its own. The draft is the average talking; the edit is where it becomes specific, true, and unmistakably yours. This session is the one the whole track has been aiming at: the human edit that separates content from slop. You will leave with a method you run every single time, not a vibe.
The anti-slop edit, step by step 6 min live
The edit is not one big rewrite; it is five specific moves in order. Cut the cliches and filler, add real specifics, sharpen the point of view, verify every fact and quote, then do a voice check. AI can assist - it will happily flag its own tells if you ask - but you make every call. The draft narrows from a generic 70% into something only you would publish.
LiveThe five moves, in order3 min▶
Run them in sequence. Cutting first clears the fog so you can see what is actually missing; specifics and point of view come next because they are what the reader remembers; fact-check protects your credibility; voice check is the last gate before it ships. Each move is a decision only you can make.
- Cut cliches + filler. Delete "in today's world", "game-changer", "unlock your potential", and every sentence that could open any article. If a line says nothing, it goes.
- Add real specifics. Replace a vague claim with a number, a name, a moment, a concrete example from your own work. Specificity is the opposite of slop.
- Sharpen the point of view. Say the thing only you would say. A draft that agrees with everyone is forgettable; a genuine take is the reason someone reads you.
- Verify every fact and quote. Check each stat, date, name, and quotation against a primary source. Anything unverified is marked and resolved, never published.
- Voice check. Read it aloud. Does it sound like you or like the model? Fix the rhythm and word choices until it is unmistakably yours.
The paragraph that flipped in four minutes. A creator ran a draft newsletter intro through the five moves: cut two cliches, swapped a vague "many teams" for the one client they actually worked with, added a single opinionated sentence, checked the one stat (it was wrong - fixed it), and read it aloud. Same length, completely different piece. The edit was the writing.
Self-studyWhy AI assists but never decides2 min read▶
The model is a genuinely useful editing assistant - it can spot its own cliches, list unverifiable claims, and suggest where a piece goes generic. What it cannot do is decide. It does not know which specific from your life is the right one, which take is honestly yours, or whether a number is actually true. So use it to surface candidates and speed the mechanical parts, and keep every judgment call in human hands. Assistant on the tells, human on the calls.
The quality bar 5 min live
"Good enough" is not a standard; it is how slop ships. You need a concrete bar you can point at, the same way every time. Five checks, each a yes/no, and the standard underneath all of them is distinctiveness: could this only have come from you? If not, it is not ready.
LiveThe five checks - each one a pass or fail3 min▶
| The check | Passes only if... |
|---|---|
| A real specific per section | Every section has at least one concrete detail - a number, name, moment - not a generality |
| A genuine point of view | It says something you actually believe, not the safe consensus everyone would write |
| Every claim verified | Each stat, quote, date, and name is checked against a source; no unverified numbers |
| Sounds like you | Read aloud, the voice is yours - the rhythm and word choices are not the model's default |
| Earns the first line | The opening makes a reader want the second line - no throat-clearing, no "in today's world" |
Self-studyWhy "fluent" is the trap2 min read▶
The dangerous thing about AI drafts is that they read fine. Grammatical, structured, confident - they pass a casual glance easily, which is exactly why unedited slop ships so often. Fluency used to be a proxy for effort and quality; now it is free, so it signals nothing. That is why the bar cannot be "is it well written". It has to be "is it distinctive and true", because those are the only signals the model cannot fake for you.
- Fluent but empty passes the eye. Slop rarely looks broken - it looks fine and says nothing. The bar has to catch "fine and forgettable".
- Distinctiveness is the real gate. Specifics, a real take, and your voice are what a reader remembers and what the model cannot supply.
Fact-checking AI content 4 min live
AI invents facts and quotes with total confidence. A plausible statistic, a real-sounding testimonial, a study that does not exist - all delivered in the same steady tone as the true things. Publish one and it is a credibility hit you do not get back. So the rule is absolute: verify against primary sources, never publish an unverified number, and mark anything you have not checked.
LiveThe verify-or-cut rule2 min▶
Treat every factual claim the model produces as unverified until you have seen the primary source yourself. Not a search summary, not "the AI said it came from" - the actual source. If you cannot verify it, you have two options: cut it, or replace it with something you can stand behind. There is no third option where you publish it and hope.
- Mark it as you draft. Tag every unverified claim with
[verify]inline so nothing slips through unnoticed. - Resolve before publish. Every
[verify]is checked against a primary source and either confirmed, corrected, or cut. A piece with a live[verify]does not ship. - Quotes are facts too. A quotation the model produced is almost never real. Attribute only what you have traced to a genuine source.
The stat that was never real. A confident "73% of teams report..." made it into a near-final post. It sounded authoritative and fit the argument perfectly - which is exactly why it was dangerous. There was no study. Caught at the fact-check gate, it became "in our own work with a dozen teams", which was true, specific, and stronger. Unverified numbers do not get the benefit of the doubt.
Run the full anti-slop pass on a draft ★ 13 min · everyone builds
Take a real AI draft - one of yours, or a Cadence piece from an earlier session - and run all five moves end to end. This is the session in one exercise: watch a generic draft become worth publishing, and time yourself so you feel how fast the edit actually is.
Pick a draft. Grab an AI draft you have, or generate a short Cadence piece (a post, a newsletter intro, a product paragraph). Keep the raw version so you can compare.
Cut. Read it once and delete every cliche, filler line, and sentence that could open any article. Be ruthless - if it says nothing, it goes.
Specify + sharpen. Replace one vague claim with a real specific from your own work, and add one sentence only you would write. This is where distinctiveness enters.
Fact-check. Tag every claim, stat, quote, and date with [verify]. Check each against a primary source. Confirm, correct, or cut - leave no [verify] live.
Voice check. Read it aloud. Fix anything that sounds like the model. Then compare against the raw draft and name the three moves that did the most work.
Build your reusable quality checklist ★ 8 min · build your own
Turn tonight's method into a gate you run every time. A short ship/slop checklist you keep next to your editor, so quality is a habit and not a mood. You draft it with AI, then edit it - the pass applied to itself.
Start from the five moves and the five quality checks. Ask AI to draft a one-page checklist with a clear "ship" line: every box must be yes before it publishes.
Make it yours. Add your formats and channels, your recurring cliches (everyone has a few), and the [verify] rule for facts. Cut anything you would not actually run.
Add the distinctiveness question at the top as the master gate: "Could only I have made this?" If a piece fails that one, the rest does not matter.
Save it where you write and run it on your next real piece. This is the gate you carry into b10, where it becomes part of Cadence's publish-readiness controls.
Try it yourself - this week ◐ 30-45 min total
- Run the full five-move pass on three real drafts this week. Notice which move you skip when you rush - that is the one to make a hard habit.
- Finalise your ship/slop checklist and pin it next to where you write. Run it on every piece for a week, no exceptions.
- Fact-check one piece properly: mark every claim [verify] and trace each to a primary source. Note how many did not hold up.
- Keep a running list of your own recurring cliches - the phrases you and the model both reach for. Add them to your "cut on sight" list.
- Optional: take a piece you published before this session and run the pass on it. The gap you find is the value the edit adds.
Official sources covered
Taught from the widely taught content-craft canon on editing and distinctiveness, plus anti-slop practice. Certificates and paid tooling stay official. This session covers:
Three questions before you go 🎯 ◐ 90 seconds
1 · What are the five moves of the anti-slop edit pass, in order?
Cut first to clear the fog, add specifics and point of view for what the reader remembers, verify to protect credibility, then a voice check as the last gate. AI assists on each; you decide.
2 · What is the single standard under the whole quality bar?
Fluency is now free, so it signals nothing. The gate is "could only I have made this?" - specifics, a real take, and your voice, the things the model cannot supply.
3 · The AI produced a confident "73% of teams report..." statistic. What do you do?
AI invents confident facts and quotes. Verify against a primary source, never publish an unverified number, and resolve every [verify] before it ships. One fake stat is a credibility hit you do not get back.