Big Idea

The argument in one line.

Trial reels only work when each version changes the opening, graphics or structure on purpose, and Claude can build those versions from a master file in one message each.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

A short-form creator with a reel that did okay and a suspicion it could do better with a different opening or pace.
Someone paying an editor $50 to $100 per video who wants cheap variations of a reel that already exists.
A creator using Claude with connected tools like Higgsfield who wants a repeatable system for testing hooks.

SKIP IF…

You don't post to Instagram, since the whole workflow is built around trial reel mechanics and Instagram's skip rate data.
You want a step-by-step tutorial on installing tools. The video shows prompts and outputs but not setup of every integration.

TL;DR

The full version, fast.

Instagram now detects reposts, so tiny tweaks like a 1.02 percent zoom or a flipped frame get flagged as duplicates. What counts is a new opening, new graphics and new structure. Alex rebuilds an editor-made reel as a master file in Claude, locks a project prompt that saves notes and shows only 10-second side-by-sides, then has Claude make five trials: a shorter straight-in cut, a 13-second micro demo, a Higgsfield-made halo opening, a motion-graphic breakdown and a black-and-white paint pour. Each test reads one metric, and a winner needs a 5 point skip rate gap.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 02:00

01 · Why trial reels need real variation

Repost detection kills tiny tweaks, so Claude is used to build new openings, graphics and structure.

02:00 – 03:10

02 · Break down the reel first

Read skip rate, saves and comments on the original to decide what each trial should fix.

03:10 – 05:00

03 · Rebuild the master file

An editor-made reel is recreated in Claude from Drive assets so it can be varied.

05:00 – 08:00

04 · Master prompt and the five decisions

A locking prompt saves notes and limits previews to 10 seconds. Seven rounds are cut to five intentional trials.

08:00 – 10:40

05 · Trial 1 and 2: straight in and micro

Remove 3.5 seconds of setup, then cut 41 seconds down to a 13-second demo.

10:40 – 14:30

06 · Trial 3: the halo

A Higgsfield-built phone orbit around the presenter tests whether motion stops the scroll.

14:30 – 16:40

07 · Trial 4: the breakdown

A frozen post with callouts tests whether spelling out the point silently beats saying it.

16:40 – 17:10

08 · A rejected experiment: scroll

Adding a hand to the scene looked wrong and was parked.

17:10 – 20:00

09 · Trial 5: the pour

A black-and-white opening floods red to test whether a gray first frame makes people wait.

20:00 – 21:58

10 · How to run the system

Limits on trials, compare only to trials, and thresholds for calling a winner.

21:58 – 23:08

11 · Vault, cohort and sign-off

Pitch for the skill vault and a 10-spot cohort, plus sponsor mention.

Atomic Insights

Lines worth screenshotting.

Instagram now flags reposts, so zooming in 1.02 percent or flipping the frame no longer counts as a new version of a reel.
A trial reel needs a new opening, new graphics and new structure to count as different.
None of your followers see trial reels, so they exist to reach non-followers and get judged in about 72 hours.
Break a reel down first: weak saves point to a saves fix, weak comments to a comments fix, and high skip rate points at the hook.
Claude can only vary a video it can rebuild, so an editor-made reel has to be recreated as a master file before variations are cheap.
Show only a 10-second side-by-side of master versus variation, because rendering the full reel every round burns credits and time.
Give the project a memory before any versions exist: save notes to Drive after every round so a long chat doesn't lose its decisions.
Cutting 3.5 seconds of setup is a legitimate trial, and in short-form that is a lot of runtime.
Cutting a 41-second reel to a 13-second demo tests whether replays and sends beat the full lesson.
Name the image model in the prompt (for example GPT Image 2.5 through Higgsfield) instead of letting Claude pick tools at random.
Tell Claude exactly what to keep: real footage stays underneath, no fake people, only the opening changes.
Many people watch reels with sound off, so a graphic opening that spells out the point tests whether it beats you saying it.
Not every experiment ships: the scroll variant with a hand in the scene looked bad and was parked after one attempt.
Limit trial reels to a few well-chosen bets on a good reel, because spamming five a day won't build a strong audience.
Compare trial reels only to other trial reels, never to normal reels, since they go to different audiences.
Call a winner only at 5 plus points of skip rate or a 1.5x gap on a per-1K number, and treat smaller gaps as noise.

Takeaway

Test one change per trial, and read one number

WHAT TO LEARN

Trial reels pay off when each version changes one element on purpose and is judged by the single metric that element should move.

Instagram flags near-duplicates, so a trial needs a new opening, new graphics or new structure rather than a small zoom or flipped frame.
Break the original down first and let weak saves, weak comments or a high skip rate decide what each trial is trying to fix.
Give the AI editor a master file and a locking prompt that saves notes, so long chats keep their decisions across rounds.
Preview only 10-second side-by-sides of master versus variation, which saves credits and time compared with rendering full reels each round.
Pair every change with a metric: opening changes read skip rate, length changes read percent watched, and ending changes read saves and comments.
Be specific about tools and constraints in prompts, such as the image model, keeping real footage underneath and changing only the opening.
Park experiments that miss after one honest try, like the scroll variant, instead of burning time polishing a weak idea.
Compare trials only to other trials, run a few deliberate ones on a decent reel, and call a winner only at a 5 point skip rate gap.

Glossary

Terms worth knowing.

Trial reel: An Instagram reel shown only to non-followers so you can test it. Instagram judges it over about 72 hours.
Skip rate: The share of viewers who swipe away in the first seconds. It is the main number for judging a new opening.
Master file: A rebuilt, editable version of the original reel that Claude can modify to create every variation.
Higgsfield MCP: A connection that lets Claude call Higgsfield's image, video and background-removal tools directly while editing.

Resources

Things they pointed at.

11:40toolHiggsfield MCP

11:40toolGPT Image 2.5

03:10toolGoogle Drive

21:58productAlex's vault and newsletter

Quotables

Lines you could clip.

01:00

“What counts now is a new opening, new graphics and new structure.”

Clear rule with no setup needed→ IG reel cold open↗ Tweet quote

21:30

“Call a winner, small gaps are noise.”

Short decision rule→ newsletter pull-quote↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

Claude's video editing has fixed trial reels forever. Trial reels are one of the fastest ways to grow on Instagram, but the hard part is making versions that are actually different. So I set up Claude once, now every new version of a reel is one message away.

New openings, new graphics, real edits I'd actually post. In this video, I'm going to show you the whole process, start to finish, five trial versions of one of my reels, all made by Claude.

Okay, so this rule right here got 10k plays, right? And I think it could perform better if I tried more variations. But the first problem is that I would have to pay my editor to create these variations on top of the $50 to $100 I'm already paying per video.

Or I'd have to spend my time, you know, creating these variations, etc. And you know, time is money. But the third problem makes this all even more complicated, which is that Instagram now spots a repost.

And you can see the days of zooming in to 1 .02 % or flip it, swap one hook line. That doesn't work anymore. It gets flagged as a duplicate.

What counts now is a new opening, new graphics and new structure. And this is exactly what Claude is able to build. And you can see Instagram have announced this themselves.

Now, today we're going to be building five different trial rules, all completely different. And you're going to see the full process right now. Now, for trial reels in 30 seconds, essentially, none of your followers see your trial reels.

The purpose of them is to attract, you know, potentially non -followers to see your content. And it's a good way to get followers and reach. They have around 72 hours to be judged, whether it's a flop, etc.

And, you know, these are the kind of numbers that I really focus on. And we'll talk more about this from skip rate to watch time to saves and sends. These are really important for us to kind of focus on because we're creating trial reels with a purpose.

And you're going to see how Claude is an absolute game changer for this. And you're going to see how Claude is an absolute game changer for this.

Now, before we do anything, you need to break down your reel. And I've built a skill for this, which will be available in my vault, but I'll be sharing more about that later in the video. Now, breaking down your reel is super important because once you break it down, then you can start spotting the opportunities to recreate that reel.

Because with trial reels, it's not just throwing it out there as many times as possible. It's actually creating trials with an intentional purpose. So you can see, for example, for the stats for this particular reel or the original reel, the skip rate.

was this, you know, the saves was pretty weak. The comments were, you know, below average, et cetera. So if I was to recreate this or create some trial versions, you know, I could potentially think, okay, how am I going to get more saves?

How am I going to get more comments? If the skip rate is super high, then I need to potentially improve the hook. I need to change certain elements.

And this is how you build a system for your content. Content systems are my number one focus. And if you want to find out more about that, hang a round.

But this is extremely important because this will spot the opportunities for you. And this alone deserves a like.

Okay, so you can see this all live in the full process. The first thing I needed to do was rebuild the master file. This particular video right here was edited by an editor, not by Claude a while ago.

Okay, so for Claude to be able to create different variations, make micro adjustments to, you know, give me those trial reels, it needs to have the master file so it can make all of that itself. So for me to build this initial version that Claude has done, I did this process right here. I had to give it these elements.

And you can see that I gave the elements in a Google Drive folder right here and then gave it essentially this prompt and, you know, all the assets. And you can see it connected. And then it was able to generate the final, you know, side by side of this reel right here.

And you can see it follows the same protocol as normal, where it gives me the beats, it asks for my permission, I reply, okay. It creates the assets, you know, it had to create the backgrounds, et cetera, using the Higgs -filled MCP, ChachiBee Image 2 .5, you know, so that it can have this kind of similarity or very identical to this video.

So it did all of that, and then finally created, you know, the cheat sheet. The Brand World color pack, which was part of my CTAs right here. And then finally, we got the output, which was very close to the original.

And I'll bring it up on the screen right now so that you can see. This is one carousel cover. Same words, same layout, same photo.

But watch what happens when the only thing I change is the color. This one feels serious, like it matters. This one feels calm, slower.

This one feels technical. distant this one feels close like a conversation nothing changed except the color same post for different feelings so the question isn't what color do you like Okay, so once that was done, so once I had my master kind of base reel, you can see that I had to give it a master prompt to begin this process of creating my trial rules.

Now, this particular master prompt is extremely important as essentially it locks everything, full project save to drive, project notes, save after every round, okay? This is extremely important because these chats can become long with all the variations, with all the changes. And, you know, before any versions, the project gets a memory so that you're not kind of losing out on things.

and you know things become messy so this is extremely important the second thing as well was that i gave it a specific way to show me the kind of variations it did for my um for trial world so for example you can see it always shows me the master plus the kind of variation that it's done side by side in a very kind of 10 second max uh format the reason for that is because every time if every variation every test is giving me the full reel every time it's going to take you ages you're going to learn through credits, and you're going to waste money and time.

So these two kind of principles here are going to make your life so much easier. Okay, so now in my first master prompt, I asked for seven rounds of different kind of versions or different trial rules that we can create. And you can see the seven rounds right here.

But kind of initially straight after that, I decided to cut this down to five as I just felt it was too much. And to be honest, five is a good amount. Like I don't think we need to abuse doing loads of different trial rules, but really being intentional with them and understanding why you're making that decision.

So you can see right here, I've gone from seven. but then decided, okay, let's flip this around and actually go with these five intentional trial rule choices. Now with your trial rules, it's extremely important to have a clean decision making system.

And I spoke about the statistics earlier in the video, you know, what am I looking at? So here are the five decisions I went for. So straight in, this is a cut.

based um kind of edit where it does a cut and you know we're going to go through all of this and what it tests is does skipping the setup keep more people you know i'm reading the skip rate the micro length based watch time plus sense this one skip rate skip rate plus percentage watch all of the trial versions that we're making they're all very intentional all testing different things you know having a feel for what the audience or what gets the most followers And this is how I've been able to come up with these decisions.

Again, this is part of a bigger content system that I'm always looking for when I'm creating my content for any social media channel. And, you know, for me, one of the beauties of using Claude editing now as opposed to potentially an editor is that we have a lot more room for testing. So you'll see in this video, you know, I'm connected to Higgs field.

So that means I can create more potential animations or use certain assets very quickly. And Claude can just integrate that super easy. It can then make adjustments based on the length, motion graphics itself.

So I have far more kind of creative freedom, believe it or not, with AI.

Okay, so the first test was trial one of five, the cut -based one, which I called the straight in. And this is a very simple one, you know, and I recommend always kind of doing this one. Essentially, this is just removing some of your intro or text, essentially making it 3 .5 seconds sooner to the payoff, okay?

Now, you might not think 3 .5 seconds is a lot, but in the world of social media, this is a lot. And essentially, what we're testing here is does skipping the setup keep more people, okay? Watch what happens when the only thing I change is the color and it's quicker, okay?

This makes a big difference for trial rules and for Claude because Claude can do this in a matter of, you know, it's a very quick change and it's something that you can get out there, you know, almost every time you do a rule. And to be honest, for me, this is one of my favorites. Sometimes we actually make scripts too long.

unnecessarily now you can see right here i followed on from the previous prompt and we went straight into this right here and you'll see here a breakdown of this prompt so you can see use line numbers line 2 and 3 to 17 no describing no guessing and you can see kind of this information that i was able to provide it this is what allows me to get an outcome like this one watch what happens when the only thing i change is the color this one feels serious like it matters this one feels calm slower this one feels technical distant this one feels close like a conversation Okay, next up we have 205, the micro.

Okay, so essentially this one is a really big change. So it's only lines three to eight. Okay, the demo, nothing else.

New headline on frame one, same post, only the color changed. So we've added this text right here. Ends on red, so it loops straight back to the start.

So this was kind of a loop back to start test. And what we're testing is, does a pure demo get more replays and sends than the lessons? So average watch time plus sends is what we're kind of looking for.

here about 13 seconds versus the original 41 seconds which is you know a massive difference so if there is a big significant like change in this one then obviously i'm getting a really good amount of data and you can see the prompt again right here h1 is approved let's move to this one And you can see kind of the outcome I got here side by side and then the prompt outcome.

So this was super clean and you can see right here the prompt behind the micro. Only the demo lines 3 to 8, about 13 seconds instead of 41. Very clear and concise, okay?

So here is what the micro actually looks like. This one feels serious, like it matters. This one feels calm, slower.

This one feels technical, distant. This one feels close. Okay, this is where I started to have a bit more fun and get a bit more experimental because I think, you know, if we're using AI, why not?

This one I called the halo. And essentially what we're looking for is that effect right there. Does that make a difference?

Does motion around my face in the first seconds stop more people? And what we're reading there is the skip rate. So let me go back.

So master 17 lines, 41 seconds. But essentially what we've added is this kind of four shot orbit right here. So there's a new opening.

Rest is kind of... this standard master file. And you can see again for the flow of that, L2 is approved.

Yes, delete the superseded notes file. And then boom, we're moving on. Now for this, we used Higgs field because we needed to generate the assets.

The idea I had here was these would be phones rather than just kind of doing carousels, having phones, you know, because carousels are essentially on social media. So therefore, if it's on social media, it connects more to the audience as it's kind of more visually, you know, realistic, I guess, in a way, you know, so. We wanted to create the assets for that.

And that was super easy. Again, in your Claude chats, all you need to do is connect it to Hicksfield. The link is in the description.

Go to MCP and you can see whatever platform you're using. You can literally connect to Hicksfield and then you'll get access to all of these features right here. And Claude can access it itself.

which is insane so you can see none of another 2 .1 you know whatever i want it to connect to it will and it can also connect for audio upscaling video etc it makes your life so much easier so you can see we created the assets so it starts creating the assets boom boom boom And you can see always something interesting as well while you're looking at your prompts.

You can always kind of just look at how Claude is, you know, actually digesting the information, coming up with your assets, with anything that it's doing. Because this is important for me, for example, I've got better at prompting at these workflows from reading, you know, what actually Claude is up to. So you can see right here, it was able to actually.

create all of them. So it created six versions of this and you can see, you know, four, four, four, it creates multiple versions. It then itself picked, you know, the best one from each one until it had these four right here.

And these were the chosen assets that it would use for this Halo. So it makes a massive difference and you can see how they were made, et cetera. What's on the screen, grow with Alex, et cetera.

And then you can see, boom, it's picked that. And this is what I mean right here. So you can see, keep the Instagram header and tab bars and then use Higgs full background remover, which is available here.

It can just literally, you know, connect to those tools, which makes Claude's life much easier because then it doesn't need to go and like, you know, find random tools to do things. And you know, these ones work as well. And it just improved these and, you know, remove the background.

So it's all ready to be used. Again, you can see right here how it starts to build everything. Super cool when you're doing it, because if you spot a mistake, well, you know, I don't necessarily recommend you watching, but if you spot a mistake, you can just put something in and say, you know, actually stop for a moment because we need to fix that.

But you can see the full float right here. And then boom, we got the Halo versus the Master. And I'm just going to play that again.

You can see right here and, you know, the kind of difference, you know, whether it makes a difference, you tell me. ultimately the data is what is going to, you know, really inform our decision. And again, just so you can see the prompts.

So use Higgs field, GBT image 2 .5. It's important to be specific always with which model you want, especially when it's for like the actual creation part, real footage underneath the moves around me, not to fake me. Text is added in the edit.

This is important. You know, you want to be clear. Okay.

So don't start adding random people or whatever, you know, keep me, get it around me. Only the opening changes and that. is the halo done.

And here is how it looks. This is one carousel cover, same words, same layout, same photo. But watch what happens when the only thing I change is the color.

This one feels serious, like it matters. This one feels calm. Next up, we had the breakdown.

Essentially, this was testing kind of Claude's motion graphics skills. If you want me to make a video about Claude's motion graphics, then let me know in the comments because all of your comments are greatly appreciated in terms of feedback, gives me information. Now, the opening freezes on the red post, so you can see.

It freezes then points at things. One of the reasons to try this is because a lot of people actually watch reels with no sound. So having something like this at the beginning, does that make someone kind of press audio on?

Does it make a difference? Again, I'm testing something. If it makes a massive difference, then maybe in my intros, I need to really kind of be particular with the way I'm showing certain, you know, assets, visuals, et cetera.

So does spelling out the point with sound off beat me saying it? And that's a massive thing that I'm testing. Skip rate plus percentage watched.

Graphics over the opening, same voice. And you can see the breakdown right here. You can see Halo approved next to the breakdown.

Show me the first 10 seconds next to the master. Wait for my OK. And you can see it started to do all of that.

Boom, the breakdown. This is one carousel. Right here.

And again, on this one, we had to make some changes. So three changes. And you can see them right there.

Go down. And essentially, the breakdown was approved. And you can see the prompt right here.

So breakdown, freeze on the red post on the phone, etc. Freeze the post, keep my face in, captions stay clear, call outs on the words. And you can see the final one right here.

This is one carousel cover, same words, same layout, same photo. But watch what happens when the only thing I change is the color. This one feels serious, like it matters.

This one feels calm, slower. This one feels technical, distant. this one feels close okay so this is where i got a bit experimental and it wasn't one of my final five trial rules but i want to show you as well that you know i create stuff that i don't always use or approve of so i tried one called scroll so essentially right here you can see okay you know carry on the scroll and essentially what i wanted to try was whether adding a hand you know to the scene etc would work whether it would look good but i didn't i to be honest i wasn't happy with the outcome And I'll play it for you right here.

It just didn't look good. And I felt like if I was to do something like this, I'd need to spend a lot more time just getting the prompt right or the vision right. But in a particular process where I'm creating these trial reels, we want to be experimental like with the Halo one.

But with this one, I think it's too experimental. And, you know, unless I had a very clear vision or a reference, then potentially I could have done it. But I think sometimes, you know, it's worth trying, but ultimately you need to make a decision, park that and move on.

But, you know, it doesn't always work. Next up, we have The Paw. And this one was one of my favorites.

I think something, again, definitely... called to try but essentially this one it opens in a black and white no color at all and the only thing i change paint pours in and floods it red you know we're talking about color something different where it goes from you know black and white and then suddenly boom you know it kind of pours the first things i had to get and you can see it kind of there right there does a gray first frame make people wait for the color especially when i'm talking about color so again i'm looking at skip rate i'm looking at watch time and you can see right here kind of the black and white paint so it's a very kind of minimal difference but not minimal i think for like, it's a visual hook.

So and what we're trying to do is identify whether that makes a difference. Now, the prompt is right here. And I'll break that down in a second.

But essentially, what we needed to do first was kind of create our assets. When creating anything a bit more, you know, dynamic, a bit more visual, you'll have to create the assets first. Again, we simply used Higgs filter that for that was able to create a bunch of different frames for my poor.

And essentially, I had my poor ready. So super happy with that. variation to his right, prove it, et cetera.

Then make, you know, you can make your tweaks like, you know, less speed, et cetera. And essentially, then we start to put it together. And to be honest, you know, I would be interested to see how this would perform.

I think maybe it could be come actually a bit quicker. I think that would have actually probably performed better. But you tell me, what do you think?

I think, you know, the idea I think actually works really well. You know, in terms, and I'll show you it in a moment, but essentially the prompt. So for paint pause pick one opening black and white paint reveals the red keep the cover sharp clean cut in line three And you know, what's some important for us to look at right here?

Is that you know, obviously the first two trial rules, you know where we're just doing cuts, you know It's a lot quicker but essentially in just maybe one or two extra prompts I'm able to get something that's completely kind of different a little bit more out there and just you know Allows us to have more variations in our content, you know I think there's nothing wrong with having that mixture of like super easy stuff but intentional stuff because it's not just adding a zoom and also adding kind of something a bit more creative a bit more experimental and you know if you have the assets available then it's amazing this is one carousel cover same word same layout same photo but watch what happens when the only thing i change is the color this one feels serious like it matters this one feels calm you know would instagram call any of these a repost in my opinion the answer is no You know, I think, you know, when you're creating your trial rules, we need to be intentional going forward.

It's not as simple as just throwing anything out there, hoping for followers. It's about being intentional. It's and, you know, ultimately the point of this particular video and, you know, these different variations, this show that you can create so many different variations.

I almost guarantee that no one in this planet would have guessed the poor. you know, the halo, et cetera. Like there's just so many different things you can do.

You know, if you put your mind to it, if you get a bit creative, AI can make your content extremely unique and extremely worth, you know, posting. You know, and I don't think any Instagram would call any of these a repost. Now there's so many different things you can test.

As I've just said, you know, you can test the hook, you can test the length, you can test the look, you can test the ending. And there's a lot more you could focus on. You know, if you want me to make videos exploring more about hooks, et cetera, again, let me know.

We're even going more into this. But essentially, there is a lot you can do. Now, here's how you run this particular system.

You know, so for trial reels, you can max five trials a day. I wouldn't necessarily do this like... every day.

I would say, you know, it's something where if you have a really good reel and you think, oh, I didn't perform that well, maybe there's a few tweaks that I could have made. That's where you start testing. I think these strategies of just spamming trial rules every day, I don't think you're going to build a strong audience that way.

I think compare trial rules to trials. I think never compare them to the normal rules because they're going out to different audiences, but it's important to kind of try to build a system where you're understanding your data. Again, once you get a winner, so if one of these performs really well, you can one of these performs really well you could push that winner within 48 hours and again you know if you changed the opening you know you're looking at potentially the skip rate if you change the length or the pace you're potentially looking at the percentage watch the watch time if you change the ending which we didn't do in this video again it might you might get more safe since comments leads All of this stuff is important information.

I hope you're not creating content for the sake of it. Call a winner, small gaps are noise. Only call it five plus points of skip rate, 1 .5X on a per 1K number.

I'll share more of that in my vault. But essentially, this is extremely important. If you want my complete system, which is going to be available in the coming weeks, it will be editing, carousels, trial rules, lead generation, everything that I do for my own business.

The link will be available in the description and there will be 10 spots per cohort. So just apply there and you'll join the wait list. And when more information is available, I will be sending it over to you.

The skill for this video will be available in my vault. All you need to do is sign up to my newsletter. You'll get access to it.

And then you can go to here and you'll see one of my latest videos or this video and you simply press open. It will be locked. You'll have to use the password that's in the description.

This is... my new system you'll see that there's other freebies if you follow me on instagram if if you're signing up to my newsletter you'll get my newsletters in there there'll be exclusive freebies that then you can also access and the passwords will be there there's also other templates etc in there as well so be sure to check it out as i'll be updating it constantly you know just part of the ecosystem but be sure to check that out now one rule is never just one test remember that I recommend watching one of these two videos.

My name is Alex. This video was sponsored by Hicksville. Be sure to check their link in the description to make your cloud editing and your workflow that much better.

My name's Alex. Let's create smarter.

The Hook

The bait, then the rug-pull.

Instagram now spots a repost, which means the old trick of zooming in a hair or swapping one hook line gets flagged as a duplicate. Alex's answer is to let Claude build genuinely different versions of a reel, one message at a time.

Frameworks