GPT 5.6 Edited This Entire Video
A YouTube tutorial on wiring ChatGPT and Claude into a $79 desktop editor called Borumi so they cut, zoom, and animate your talking-head videos for you.
August 19thA three-document Claude Project, a rambled voice memo, and a locked ten-slide template replace generic AI prompting with carousels that sound like one specific person.
Generic AI output comes from generic prompts, not from AI itself, so loading a Claude Project with a documented avatar, tone of voice, and brand book, then feeding it a rambled voice note, produces carousel copy that sounds like the creator instead of a chatbot.
The creator's carousel system starts with a Claude Project loaded with three documents: an avatar doc describing one ideal customer, a tone-of-voice doc built from real writing samples, and a brand book with her founder story. Each week she rambles into a voice memo for about fifteen minutes, feeds the transcript to Claude to surface themes, then turns one idea into a hook and a fixed ten-slide carousel (hook, problem, framework, proof, bridge, CTA). She edits only by reading the draft aloud and cutting anything she wouldn't actually say, reuses the last slide as the caption, and tracks revenue and leads instead of views. She calls the non-negotiable business layer the Rule of Three Ones: one high-ticket offer, one customer path, one metric to measure.
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 →
She states her Instagram business results and the promise that Claude, not AI slop, is now her carousel engine.

A single carousel made with this system: 1,050 new followers, 5,200 profile visits, 83% non-follower reach, 1,300+ comments, from about 15 minutes of work.

Her first attempts using generic AI chat produced flat, generic drafts until she realized the fix was setting up a dedicated Claude Project for her content.

Open the Claude desktop app, create a new Project, and prepare to load it with three reference documents.

Avatar doc (one named ideal customer), Tone of Voice doc (real writing samples, including deeply personal ones), and Brand Book (founder story plus visual identity) — each takes 30-45 minutes and gets reused in every future chat.

A 10-20 minute unscripted voice memo (recorded on a ferry with no service, in her example) gets transcribed and fed to Claude with the prompt 'what themes or patterns are you seeing' to surface content ideas.

One idea from the ramble becomes a hook, written by the human but refined with Claude by requesting variations tuned by constraint (more controversial, more specific numbers, no hype language).

A fixed prompt and slide map: slide 1 hook, slides 2-3 problem, slides 4-7 framework/methodology, slide 8 proof, slide 9 bridge, slide 10 CTA.

Edit only by reading the draft out loud and cutting anything that doesn't sound like something she'd actually say; keep the slide design simple; the caption is just the final slide's text, copied and pasted.

Following the system dropped her per-post time from hours to about ten minutes once it became repeatable.

The business layer AI can't do for you: one high-ticket offer, one customer path every post points to, and one metric to measure — revenue and leads, not views.
Generic AI drafts come from generic prompts, and the fix is loading a persistent AI workspace with real proof of how you think, talk, and sell before you ever ask it to write.
“Why even use AI in the first place if I can just do it better?”
“The hook is where the money lives.”
“If you're feeling like it's AI slop, promise you other people will too.”
“Forget views, followers, and likes. It's vanity. You need to track revenue.”
“Without the Rule of Three Ones, you're gonna have views but no business.”
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.
She built a seven-figure Instagram business without AI, then watched her first ChatGPT drafts come back sounding like everyone else's. The fix wasn't a better prompt: it was giving Claude three documents and one rambled voice memo before ever asking it to write.
Loading these three documents into a persistent Claude Project means every new chat already knows the writer's voice, offer, and buyer instead of starting from a blank prompt.
A fixed slide-by-slide map used as the prompt structure for every carousel, so writing one becomes filling in a template instead of solving a blank page.
Speak unscripted into a voice memo for 10-20 minutes about whatever is on your mind, transcribe it, and ask Claude what themes or patterns it sees rather than writing from an empty prompt. Produces multiple separate content ideas per session rather than one polished piece.
The business constraints AI tools can't set for you; without them, content gets views but doesn't build a business.
“Watch it next here.”
Soft close pointing to another of her videos rather than a hard sell on camera; the actual lead magnet (the carousel playbook) is only pitched via the video description link, not spoken.
00:00
00:13
00:19
00:30
00:39
00:46
00:51
01:04
01:10
01:17
01:26
01:43
01:51
01:54
02:03
02:10
02:26
02:31
02:35
02:46
02:53
03:05
03:17
03:22
03:29
03:42
03:46
03:56
04:07
04:14
04:19
04:29
04:36
04:46
04:56
05:06
05:11
05:23
05:31
05:37
05:51
06:00
06:06
06:16
06:23
06:33
06:41
06:49
06:58
07:02
07:17
07:21
07:32
07:41
07:46
07:55
08:07
08:15
08:24
08:32
08:41
08:46
09:02
09:04
09:12
09:22
09:37
09:43
09:48
09:55
10:03
10:16
10:25
10:33
10:38
10:47
10:57
11:06
11:19
11:25Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
A YouTube tutorial on wiring ChatGPT and Claude into a $79 desktop editor called Borumi so they cut, zoom, and animate your talking-head videos for you.
August 19thA three-stage diagnostic for why your Instagram Reels aren't selling — and the exact fix for each stage.
August 4th 2025A rebuild of a two-year-old ChatGPT workflow, now wired through Claude's native Canva integration — most of the runtime is the unglamorous setup work that has to exist before one-prompt batching works at all.
July 8thA 100x Engineers host walks through the exact Claude Projects setup he uses to auto-generate Instagram Reel scripts from 400 of his own past posts.
November 11th 2024A six-rule prompting cheat sheet distilled from Anthropic's own best-practices doc for the model creators internally call Fable 5.
July 1stA 5-minute setup tutorial that turns Claude Desktop into a cinematic AI video studio by wiring it to Magnific via MCP.
June 23rd