Personal brand plan · 21 October 2026

Be the person people come to for AI that does real work.

Built from six research sweeps: the top AI creators on Instagram, TikTok, YouTube, X and LinkedIn over the last 90 days, what their best posts did, why it worked, and how they turn viewers into buyers. Plus audience psychology and what each platform's ranking system rewards.

“I run my company on AI employees, and I build them for real businesses. I show the receipts.”

It is what wins right now

October's top Instagram AI reels are AI-employee proof: 1.85M plays for an agent that caught 43 failed payments overnight; 528K from a 38K account saying “I'm the only human in my company”.

Nobody owns it

None of the top 30 AI YouTubers installs AI employees in real local businesses on camera. On LinkedIn, the “AI employee” story is being paid for by a product. Your unsponsored proof stands out.

It sells what you sell

The loudest complaint in 1,982 real comments is hype with no real costs. Receipts answer it, and owners with budgets sit on LinkedIn and YouTube.

Your question

Good content, or content that gets views?

Neither alone. Depth is the subject, reach is the packaging. Every post teaches one real, checkable thing, framed so a business owner gets it in three seconds. Views tell you which topics to make; owner DMs and booked calls tell you what to sell.

0 sales

from Hormozi's six most-viewed videos last quarter (his own words, 25 Sep 2026). His best-selling videos had 100K to 270K views.

11.7 vs 3.2

paid community members per 1,000 YouTube subscribers: Jack Roberts (287K subs) vs Nate Herk (1.09M). Specific beats big.

15M views, 42 comments

on a sponsored AI Short, vs 47K views and 2,479 comments on one specific keyword reel (aryan.mjn). Spectacle brings no buyers.

Your weekly mix

First 90 days (until about 5-10K followers)

Reach 60%
Proof 25%
Human 15%
Reach 60%Proof 25%Human 15%

After that (Jun Yuh's 3-1-2-1, adapted)

3 reach
2 proof
1 human
1 convert
3 reach2 proof1 human1 convert

Reach = AI news and demos in business terms. Proof = real installs, costs, results. Human = your story. Convert = a free resource or a call.

About Dhruv's approach (“post what people want to hear”): it does grow followers fast. His likely account has a 1.68M reel on about 20K followers. Hormozi's data says those followers want more of the same, not your product. Use it as the reach lane, only when it speaks to the same business owner.

Three rules

What to talk about

Seven pillars, all from work you already do

1

My AI employees

The fleet that runs your company: names, jobs, what they did last night, what they cost.

Flagship: proof + reach
2

Installs in real businesses

Front desk, missed calls, lead follow-up, product photos. Rebuilds with a deadline. Cost per job.

Brings buyers
3

AI news in business terms

Within 48 hours of a launch: test it on a real job and say what it means for an owner.

Reach
4

The honest numbers

Costs, hours, what stuck, what got fixed, your real opinions.

Trust
5

Copy this, free

Job descriptions, the 4-question scope test, checklists, prompts, behind a comment keyword or a link.

Convert
6

AI creative

Cherry Shot before/afters and AI-made product ads.

Reach + product
7

The human side

Building in Vancouver and India; true moments from client visits.

Follows
+

India lane (test)

Hinglish versions of pillars 2 and 5 as Trial Reels from week 4. One Hinglish AI reel got 474K plays and 17,155 comments.

Test
Leave out: side-hustle income claims, AI realism spectacle, paid-partner news, reposts, fake people (your twin is you, labelled as AI), comment bait and engagement pods. They win views and attract people who will never buy, or they break trust.
Platform playbooks

Each platform has a different job

Instagram brings reach, LinkedIn brings buyers, X brings credibility with builders, YouTube comes in phase 2, and TikTok gets free cross-posts.

Reach engine. 5 reels a week, every one cross-posted to TikTok.

Who is there, and their mood

Young owners, side-hustlers, students and job-seekers (India has the largest Instagram audience in the world). Mood: entertainment with a "teach me something quick" slice. 79% of US adults aged 18-29 use it (Pew 2026).

What the platform rewards

Shares to DMs per view, watch time, likes per view, originality (accounts that mostly repost stop being recommended), recency. Trial Reels test a topic on non-followers first.

What is winning (last 90 days)

  • AI-employee proof, almost no words: lukebuildsai 1.85M, jakedavidsai 528K from 38K followers
  • Split-screen "it's called X and it's free", small face at the bottom: opusjake 2.16M, nick_saraev 467K
  • Numbered map or list: sabrina_ramonov "10 secret codes" 1.8M, Jarvis 9-step map 1.57M
  • Before/after and "AI made this" reveals: damianodesu 1.53M from 6K followers
  • Reels over 300K plays run about 60 seconds; long is fine if every beat changes

Opening lines that worked (verbatim)

  • “43 payments failed overnight and I found out from Jarvis on my way to dinner with my girlfriend.”
  • “I'm 20, I play college football, and I'm the only human in my company.”
  • “You can now run Claude Code completely free with almost unlimited usage.”
  • “Your dashboard is lying to you. It says you made $1,284 today. You actually kept $412.”

Turning viewers into buyers

"Comment KEYWORD and I'll DM it to you" at the end. Comment rate 1.03% with a keyword vs 0.09% without (3.8x inside the same account). It brings buyers, not reach: only 30% of the top 40 reels have one.

Gurleen's first posts here: Topics 1, 3, 7, 9, 37, 43 in the topic bank below.

Top creators right now

#CreatorFollowersMedian viewsWhat they are the go-to forHow they make money
1Evolving AI
@evolving.ai
4.83M609KAI news clips and AI-video viralityNewsletter 100K+ readers; sponsorships (inferred)
2Varun Mayya
@thevarunmayya
1.2M403KIndia's best-known AI/tech creator: daily AI-news reelsAeos Group / Yaas Media agencies, video-editing school, game studio, YouTube 1.18M
3Trakin AI (Arun Prabhudesai)
@trakin.ai
998K193KHinglish AI tool/prompt reels for India ('Fastest Growing AI Instagram Page')prompts.trakin.ai prompt library; Trakin Tech brand
4Ankita
@creativewithankita100
41K169KAI/Replit side-hustle tool reels, 15 secondsunknown
5iinsightri
@iinsightri
70K121KAI realism / Higgsfield Genjutsu skits ('everything you scroll past might be AI')Brand collabs (Higgsfield); free workflow doc as lead magnet
6Nick Saraev
@nick_saraev
694K118KFree Claude Code tools, MCPs and AI automation for beginners chasing a first AI clientMaker School on Skool $184/mo (self-reported '$160K/m across two agencies'); LeftClick AI agency
7Chandler (Chanyoung Ryu)
@chandlerintelligence
95K106KClaude and agent teams for creators ('team of friends' bots)not found
8TechPlus AI (Naman Deshmukh)
@techplus.ai
1.3M92KHinglish AI explainers under 60 secondsYouTube, sponsor-run page (@socialsauce.in)
9Riley Brown
@realrileybrown
332K57KAgent-native / Claude Code and Codex demos (Vibecode)Vibecode app; sponsors
10Forgemind AI
@forgemind_ai
63K45KAI agents/automation training (India), 'Official Meta Business Partner'10K+ students trained; Live Lab classes; Corporate training
11Peter Griffin AI
@petergriffin.ai
164K37KClaude skills/plugins lists voiced by Peter Griffin + StewieCollabs email; beacons.ai link; Telegram broadcast
12Luke Cutting
@lukebuildsai
283K32K'My AI agents run my business while I sleep' (Jarvis, Azaris)Azaris (AI agents product, launch.azaris.ai); free email course

Top posts (one per creator, paid-looking posts removed)

ViewsCreatorOpening line (verbatim)x own median
6.23M@iinsightriEverything you scroll past might be AI now - ads, UGC, creators, all of it.51.4x
5.82M@chandlerintelligenceBeing a content creator can get lonely at times. That's why I created a team of friends to work with.54.9x
4.87M@evolving.aiA Tesla Model S suffered a tire blowout at around 177 mph (285 kph) on Germany's Autobahn.8.0x
3.47M@kayvon.aidear bruce...254.8x
2.16M@opusjakeIf Claude keeps giving you bad answers, you need to make 100 versions of Claude fight to the death.118.5x
1.85M@lukebuildsai43 payments failed overnight and I found out from Jarvis on my way to dinner with my girlfriend.53.7x
1.8M@sabrina_ramonov10 secret codes to learn anything. (on-screen: 'Secret Codes to Learn Anything')59.9x
1.68M@maujawaan.techkeep your moments secure with this quick and easy check826.2x

Full data: creators.csv · posts.csv

Buyer engine. 4 posts a week, written from real work, one image each.

Who is there, and their mood

Owners, executives and managers with budgets. 45% of US households earning $100K+ use it vs 18% under $30K (Pew 2026). Mood: career and status, "make me look smart at work".

What the platform rewards

Time spent reading, comments with substance, saves and shares, match with the reader's job. LinkedIn is killing engagement pods, comment automation and "comment YES" bait (Tim Jurka, 2026-03-12).

What is winning (last 90 days)

  • Decision frameworks for whoever signs the spend: Alex Wang's "don't start with an agent" test, 20.8x his median, 754 reposts
  • Launch-day recipe tested on real inputs: Charlie Hills, Claude Motion, 3,210
  • Short true personal stories make the biggest spikes: Ishan Sharma 13,621
  • Text + one image, 200-300 words. Number-led hooks 1.16x vs question hooks 0.73x. Link posts 0.58x
  • Big following does not mean reach: Zain Kahn 1M followers, median 34 reactions; Charlie Hills 200K, median 926

Opening lines that worked (verbatim)

  • “The fastest way to make an AI project expensive is to start by saying: "Let's build an agent."”
  • “The best thing you can do for your AI strategy in 2026 has nothing to do with AI.”
  • “I gave Claude Code access to a business of mine that does around $1M a year and asked it to automate everything.”
  • “When someone shows me a new AI agent, I ask one question before the demo starts. How do we turn it off?”

Turning viewers into buyers

A repost ask lifts a post to 1.22x the creator's median. Put links in the first comment. Model to copy: Andreas Horn, who writes to executives and sells a $2,499 cohort plus a 38K-subscriber newsletter. Watch out: Viktor (an AI coworker product) is paying creators for #AIemployee posts, so unsponsored proof is the difference.

Gurleen's first posts here: Topics 4, 13, 14, 15, 16, 36 in the topic bank below.

Top creators right now

#CreatorFollowersMedian engagementWhat they are the go-to forHow they make money
1Pascal Bornet
pascalbornet
2M1.55KAI + automation thought leadership; agentic AI for enterprisesSponsored partner posts (#SalesforcePartner, #WorkdayPartner, #BCGPartner, #sponsored) are 7 of 20 posts and his sponsored median (1,916) beats the rest (1,107); keynotes; book IRREPLACEABLE; irreplaceable.ai
2Charlie Hills
charlie-hills
200K926Practical AI for creators/marketers: Claude, infographics, motion; AI content systemsAI OS for LinkedIn: 6-week cohort, founding cohort $1,500, starts 2026-11-09 (charliehills.ai/aios); Substack newsletter; HubSpot collaboration
3Ruben Hassid
ruben-hassid
831K909Claude/AI how-to for working professionals (How to AI)How to AI Substack (free + $200/yr paid, "thousands of paid subscribers" per Substack; claims 397,034 weekly subscribers in a post URL), enterprise Claude consulting (self-described), EasyGen tool
4Dharmesh Shah
dharmesh
n/a813HubSpot founder: AI agents, startup product launches, dinners/communityHubSpot; YouSpot personal CRM (discounted $8/month link offered in his top post); LinkedIn newsletter
5Ishan Sharma
ishansharma7390
519K590India YouTuber (2M+): upskilling, business, AI; hackathonsEnterprise AI trainings, MarkitUp media, hackathons (MumbAI Hack Day), hiring posts
6Mohini S.
mohini-sss
n/a571India AI tools for marketers and corporate professionalsAI & marketing consultant ("DM for collaboration"); promotional posts with product links (Abacus.AI ChatLLM; sponsor or affiliate relationship unconfirmed)
7Greg Coquillo
greg-coquillo
n/a515AI infrastructure explainers from an AI platform product leaderEmployer role (Azure AI infrastructure); paid partnership (Viktor, #AIemployee)
8Ethan Mollick
emollick
439K490Research-backed AI at work and school (Wharton)Books (Co-Intelligence NYT bestseller; Co-Existence released 2026-10-20), Wharton role, speaking
9Allie K. Miller
alliekmiller
2M408AI for business leaders: enterprise adoption, agents, what to build firstAI-First Academy course (price not confirmed; she claims 400K+ students), free 5-day AI Fast Track email course, AI with Allie newsletter (claims 2M readers), Open Machine (advisory), keynotes
10Vaibhav Sisinty
vaibhavsisinty
604K392India AI upskilling: GrowthSchool workshopsGrowthSchool: free 3-hour AI workshop upsells; entry workshops Rs 499-2,998 (third party); says $6M to $25M+ scale and 1M learners (self-reported), raised $5M
11Anisha Jain
theanishajain
193K389How to (actually) use AI at work; LinkedIn growth with AI (How To AI, with Ruben Hassid)Newsletter; paid partnership (Viktor, #AIemployee); comment-keyword lead magnets ("Comment GROWTH")
12Alex Wang
alexwang2911
1M326Learn AI Together: practical AI explainers, free AI community, newsletterSponsored posts: 13 of 20 posts carry promoted events/trials (DigitalOcean+NVIDIA Open Intelligence Summit, Oracle AI World, NVIDIA links, free-trial links); LinkedIn services page; newsletter

Top posts (one per creator, paid-looking posts removed)

EngagementCreatorOpening line (verbatim)x own median
14Kishansharma7390Had the most emotional day of my life. / Finally got this for my parents.23.1x
6.79Kalexwang2911The fastest way to make an AI project expensive is to start by saying: / “Let’s build an agent.”20.8x
5.23Kpascalbornet𝗬𝗼𝘂𝗿 𝗲𝘅𝗽𝗲𝗻𝘀𝗶𝘃𝗲 𝗴𝗶𝗺𝗯𝗮𝗹 𝗵𝗮𝘀 𝗯𝗲𝗲𝗻 𝗼𝘂𝘁𝗽𝗲𝗿𝗳𝗼𝗿𝗺𝗲𝗱 𝗯𝘆 𝗮 𝗰𝗵𝗶𝗰𝗸𝗲𝗻. / Somewhere, an engineer has spent three years refining gyroscopes, motion sensors, and stabilization algorit...3.4x
3.59KdharmeshFor the past 30+ years, I've been hosting informal, last-minute dinners in whatever city I'm traveling into. I've had the chance to make some great connections with so...4.4x
3.21Kcharlie-hillsAnthropic (finally) dropped Claude Motion. / I made 16 Apple-style animations. Here's how:3.5x
3.17KvaibhavsisintyIt was super cool to meet the AI boss 🚀 / I had a ton of questions lined up. Got a chance to ask a few of those.8.1x
2.99Kruben-hassidGoogle your name. Now imagine your ex doing it. / Four prompts erase what they'd find (for $0):3.3x
2.82Kandreashorn1The best thing you can do for your AI strategy in 2026 has nothing to do with AI. / It is fixing your data.8.7x

Full data: creators.csv · posts.csv

Credibility with builders and founders. One Article plus 3-4 short posts a week.

Who is there, and their mood

18% of US adults, male-leaning, slightly higher income. Developers, founders, agencies, investors. Mood: insider news and status, "what is happening in AI right now".

What the platform rewards

X's open-source ranker (Aug 2026) adds up predicted replies, reposts, DM shares, reading time, profile clicks and follows, minus mutes, blocks and reports. Replies that the author answers weigh the most. Small accounts get a new-author boost.

What is winning (last 90 days)

  • X Articles (long posts) with "full guide" titles: 2.4x an account's median, 42% reach 5x or more
  • Small accounts hit big with Articles: Sabrina_Ramonov 498K from 7.6K followers; Argona0x 1.87M from 37K naming each AI employee and its job
  • Posts with dollar figures are 11% of posts but 18% of the 300 most viewed
  • Reaction openers ("This is wild") 1.93x; quote posts are weakest

Opening lines that worked (verbatim)

  • “Grok Bot built me a completely autonomous business which has already generated $1,100”
  • “6 months ago: laptop open all day, typing into Claude Code, waiting.”
  • “Today is a special day for me. I've made $3M with my 36 startups.”
  • “When you sell your business, never ever agree on an earn out”

Turning viewers into buyers

Nobody sells on X itself: the link goes in the first self-reply (YouTube, newsletter, booking page). Answer replies in the first hour; Trejon drafts them into the same approval batch. Starting point: @gurleenAI has 5 followers and a best post of 1,111 views.

Gurleen's first posts here: Topics 2, 6, 8, 17, 28, 34 in the topic bank below.

Top creators right now

#CreatorFollowersMedian viewsWhat they are the go-to forHow they make money
1Logan Kilpatrick
@OfficialLoganK
383K296KGemini / Google AI Studio launches as they happen (employee voice, not a creator business)None personal. Employer distribution for Gemini API / AI Studio [P: bio]
2Pieter Levels
@levelsio
980K141KSolo founder shipping AI products in public with live revenue in the bioBio lists live monthly revenue per product: Photo AI $86K/m, then $44K/m, $23K/m, $13K/m (next to "@X"), $7K/m, $5K/m [S: bio 2026-10-21]; his own post says business revenue stable at $200-250K/mo [S: x.com/levelsio/status/2102739359715774665]; MAKE book 32,168 copies / $955,903 [S: readmake.com via startupswiki]
3Alex Finn
@AlexFinn
485K135KAutonomous-agent / vibe-coding demos for non-engineers (Grok Bot, ChatGPT Voice, local models)Vibe Coding Academy on Skool: 681 members at $99/month [P: skool.com/vibe-coding-academy/about 2026-10-21]; Creator Buddy "$300K ARR" [S: bio]; Henry Intelligent Machines
4Greg Isenberg
@gregisenberg
748K121KStartup ideas and AI-native business models; podcast host; Late Checkout CEOLate Checkout (builds companies: Ideabrowser, others), Startup Ideas Podcast, sponsors [S: bio]; prices not captured
5Marc Lou
@marclou
415K80KIndie hacker: ships tiny startups fast, shows every revenue numberHis own page: $94.2K/month across startups: TrustMRR $50.8K/mo, DataFast $31.5K/mo, CodeFast $5.7K/mo, ShipFast $1.1K/mo; newsletter 50K readers [S+P: marclou.com 2026-10-21]; ShipFast peaked $133K/mo Apr 2024 [S: newsletter 2026-09-08]; his own X post: "ShipFast in 2026: $2000/mo"
6Matt Shumer
@mattshumer_
398K59KEarly-access hands-on takes on new models; viral demo videosInvestor; "Something Big" newsletter sells ad spots (self-reply: "more ad spots to sell, sponsor@somethingbig.ai") [P: thread 2095596175705399482]
7McKay Wrigley
@mckaywrigley
229K59KBuilds and teaches AI coding workflowsNot confirmed (site mckaywrigley.com not fetched)
8Chubby
@kimmonismus
152K59KFast AI news analyst; model-release and benchmark coverage; Superintel editorEditor-in-Chief of Superintel, "community of 300k+ in total" [S: bio]
9Amjad Masad
@amasad
500K56KReplit CEO: AI coding agents and softwareNone personal; Replit distribution [P: bio]
10Ethan Mollick
@emollick
407K51KWharton professor: research and tests on AI at workBook "Co-Existence", Substack [S: bio]
11Min Choi
@minchoi
392K46KWhat is wild and what is practical in AI: demos and workflow listsNo product link in bio; sponsorship not confirmed
12Rowan Cheung
@rowancheung
601K44KDaily AI news; founder of The Rundown AI newsletterNewsletter is free and ad-funded; advertiser page claims 1.7M on the daily AI newsletter and 2.5M+ across three [T/S: builtbyfoundry.io 2026-10-06 citing advertiser page]; Rundown Pro $29/mo (was $84/mo on 2026-08-27), "70% expense it" [S]

Top posts (one per creator, paid-looking posts removed)

ViewsCreatorOpening line (verbatim)x own median
18M@VaibhavSisintyDyson trained an AI on 470,000 dental images and put it inside a toothbrush. With a camera. That flosses for you. In real time.1886.6x
12M@mattshumer_My first "holy shit" moment with GPT-6 Astra:198.9x
7.91M@minchoiThis is literally my new workflow now:172.6x
7.7M@aaditshSo let me get this straight…439.8x
6.02M@amasad18x improvement in intelligence per joule in 16 months. https://t.co/lP62kI5rpr107.5x
5.49M@AlexFinnI don’t think you understand what ChatGPT Voice unlocks40.7x
5.29M@LinusEkenstamA friend in Sweden just asked me what skiing gear to get for this season…481.5x
5.17M@deanwperkins16 minute tutorial on how to build $50,000 cinematic websites (step by step) with Claude Fable 5279.0x

Full data: creators.csv · posts.csv

Highest-intent buyers. Phase 2: one long video every two weeks, only if Gurleen records.

Who is there, and their mood

84% of US adults, 90% of $100K+ households. Mood: search and learn, willing to sit 20-60 minutes. The most trust per viewer of any platform.

What the platform rewards

Viewer satisfaction: does the video keep the title's promise. Then relative watch time, click-through on the thumbnail and title, and viewers coming back.

What is winning (last 90 days)

  • First-person experiments (1.4x) and business or money titles (1.5x) make the surprise hits; listicles 0.65x
  • Rebuild with a deadline: Liam Ottley's accounting-firm rebuild in 4 days, 4.8x his median
  • News works only on day 0-1 after a launch (3.2x), then fades to 1.0x
  • Fewer uploads earn more views per subscriber (under 1 a week: 0.40; over 4 a week: 0.10)
  • Shorts are not the growth engine: median 8.2K vs 42.8K for long videos

Opening lines that worked (verbatim)

  • “I Rebuilt This 46-Person Accounting Firm With AI (in 4 Days)”
  • “I Tested The Viral Github Repo That Reverse Engineers Anything”
  • “Why I Cancelled My Claude Code Subscription”
  • “I Spent $21,455 & 1,200 Hours On Claude Code & Codex For This”

Turning viewers into buyers

Description line 1: the offer or call. Line 2: the free resource. Mid-size channels convert best: Jack Roberts 11.7 paid members per 1,000 subscribers vs Nate Herk 3.2. Thumbnails: real face plus 1-4 big words plus a logo, or a proof object (a dashboard, a payment). Left out of the table: Shark Numbers (1.64M median), whose sponsor-funded reviews show an engagement pattern that looks paid.

Gurleen's first posts here: Topics 10, 11, 23, 29, 39 in the topic bank below.

Top creators right now

#CreatorFollowersMedian viewsWhat they are the go-to forHow they make money
1Fireship
@fireship
4.31M1.62MDev/tech news in 5-7 minute dense, joke-heavy videos; about half the titles are AIbytes.dev newsletter + fireship.dev (courses) in every description; sponsors (CodeRabbit, Railway, Blacksmith)
2AI Search
@theaisearch
764K321KModel-launch and research explainers, 25-33 minute analysis with curiosity titles ("GPT 6 Astra is a freak")Substack newsletter (aisearch.substack.com), Ko-fi, hardware/sponsor links (Nvidia, Dell) in every description
3AI Explained
@aiexplained-official
456K237KFrontier-model news and analysis for people who follow the researchPatreon (linked in all 7 descriptions), newsletter (signaltonoise.beehiiv.com) and podcast; prices not checked
4Tina Huang
@tinahuang1
1.36M222KAI + data for careers, "X in 24 minutes" explainersLonely Octopus (AI Agent Bootcamp cohorts, waitlist), free 28-Day AI Sprint Roadmap, sponsors (Crusoe, 365DataScience, StrataScratch)
5Varun Mayya
@varunmayya
1.18M186KIndia tech/business thinker, long interviews and masterclassesNot visible in descriptions (only a Build What Moves India event page). Unconfirmed
6Jeff Su
@jeffsu
1.93M179KProductivity + AI workflows for office professionalsFree newsletter + templates, paid academy courses (academy.jeffsu.org, coworkacademy.ai), sponsors (HubSpot, Granola); prices not checked
7Greg Isenberg
@gregisenberg
732K162KStartup ideas + building with AI; interview formatLate Checkout agency, Idea Browser (paid idea-research product, pricing page not retrievable), Startup Ideas Podcast, thevibemarketer.com
8Two Minute Papers
@twominutepapers
1.84M160KResearch-paper news in ~5 minute narrated videosSponsor (Lambda in 21 of 24 descriptions)
9Vaibhav Sisinty
@vaibhavsisinty
873K141KIndia AI news digests ("+21 AI updates") and AI-income ideasFounder of GrowthSchool (per a third-party testimonial page, not confirmed on his channel); recurring links join.switchit.app (57/90 descriptions, not opened) and a StayingAhead newsletter short link
10Matt Wolfe
@mreflow
1.02M135KAI news and tool discovery (Matt Wolfe)FutureTools.io directory/newsletter (linked in 48 of 65 descriptions), sponsors, podcast
11Wes Roth
@wesroth
332K134KAI news commentary, ALL-CAPS curiosity titlesNewsletter (natural20.beehiiv.com in 39 of 40 descriptions); no course/community found
12Jay E | RoboNuggets
@robonuggets
203K122KClaude/agent tool roundups and tutorials for buildersSkool paid community $97/mo (1,445 members) + free Skool network (98,306 members); affiliates (ElevenLabs, Apify, n8n, Make, Blotato)

Top posts (one per creator, paid-looking posts removed)

ViewsCreatorOpening line (verbatim)x own median
7.29M@marktilburyI Tried The LAZIEST Way to Make Money With AI || FIRST WORDS: "People are secretly making thousands of dollars using Claude to make AI influencers like this, this, and...3.9x
6.46M@fireshipEvery app on Earth is now open source... whether it likes it or not || FIRST WORDS: "Last week, a developer got so tired of Adobe's predatory business model that he de...4.0x
2.44M@futc.photographySomebody Vibe Coded EVERY SINGLE ADOBE App20.6x
2.24M@sandeepswadia4 AI Agents To Automate 99% Of Your Life || FIRST WORDS: "four AI agents you can build today that will save you hours of work every week. You don't need any technical ...3.7x
1.88M@ayushmanpanditaReal vs AI Challenge (she almost won!)104.0x
1.88M@ivankvI Stopped Testing Features and Gave Manus a Real Project – Manus AI Review || FIRST WORDS: "The app is done. I'll enter $50 as the hourly rate and click get started. >...1.1x
1.82M@siliconvalleygirlAndrew Ng: The Biggest Opportunities in AI Aren't Where You Think18.2x
1.73M@meticsmediaThe Ultimate Beginner’s Guide to Claude AI || FIRST WORDS: "Hey, I'm Matt and this is the ultimate beginner's guide to Claude AI. By the end, you'll have it fully mast...130.1x

Full data: creators.csv · posts.csv

Free extra reach. Cross-post every Instagram reel; no separate production yet.

Who is there, and their mood

65% of US adults aged 18-29, lower-income skew. Mood: scroll and escape, but news-hungry (53% of users get news there, Pew 2026).

What the platform rewards

Watch time and completion, shares, comments, follows from the video. Slow openings die.

What is winning (last 90 days)

  • Cross-posted educator reels get 0.1-0.4x of their Instagram plays
  • Organic winners are personality POVs (gazi.ai 1.4M, 26,700 shares) and save-worthy tutorials (aisavvy 18,451 saves)
  • The biggest AI numbers are paid partner posts (Kane Kallaway 15.2M for Google)

Opening lines that worked (verbatim)

  • “This video is 100% edited by AI. It did every cut, motion graphic and texture you're seeing right now.”
  • “You can now give Claude Code a phone number in just five minutes.”
  • “I'm giving myself 90 days to make a hundred grand with AI.”

Turning viewers into buyers

There is no comment-to-DM tool here, so swap the call to action to "link in bio" when cross-posting.

Gurleen's first posts here: Cross-post everything; test topic 50 as a native POV in the topic bank below.

Top creators right now

#CreatorFollowersMedian viewsWhat they are the go-to forHow they make money
1Gazi Jarin
@gazi.ai
668K79KSoftware-engineer lifestyle POVs, coding and apps (personality-led, not pure AI)brand ads (#ad), beacons link
2Kallaway
@kanekallaway
447K32KTech and AI news explainers for creatorsBuilding sandcastles.ai; AI-company partnerships (#googlepartner, #ChatGPT_Partner)
3Riley Brown
@rileybrown.ai
638K21KAgent tools and Claude Code demos (vibe coding)Vibecode app; sponsors
4BioMakers AI
@bio_makers1
6M14KHumanoid robots and applied AI (Spanish)BioMakers AI (linktr.ee)
5Rowan Cheung
@rowancheung
80K11KAI and robotics news (The Rundown AI)The Rundown AI newsletter 2M+, Rundown University
6Sabrina Ramonov 🍄
@sabrina_ramonov
862K6.58KChatGPT/Claude commands and automations ('teach 10M people AI')Blotato from $29/mo; free courses link in bio (Sabrina.dev)
7Justin Fineberg
@justinfineberg
288K5.13KAI agents that operate like your team (Cassidy.ai CEO)Cassidy.ai
8Fawzi
@fawziammache
134K4.89KExplaining AI without the hype with hand-drawn visualsfree AI explainers library
9Victor C
@aisavvy
134K4.78KAI tools and tutorials (Claude + Blender + Seedance)aisavvy.net, collabs
10AI Century
@aicenturyclips
430K4.35KAI clips and viral AI videosbio link 'learn the secret to make viral clips'
11Jo Mendes
@nomadatoast
339K4.29KCreative tech tips and AI tools (Higgsfield API)Higgsfield partner; liinks.co
12Tommy Teja Adhiraja
@tommythings
1.3M4.08KAI for entrepreneurs (Indonesia, English)AICO Community, Zando Agency, Mechalens AI

Top posts (one per creator, paid-looking posts removed)

ViewsCreatorOpening line (verbatim)x own median
1.4M@gazi.ailink dropping tomorrow 👧 #povreels #workmeetings #college #coding [caption; audio is music]17.7x
369K@aisavvyOne of the best ways to control character and camera movement in your AI videos is to create a scene in Blender first.77.2x
354K@rileybrown.aiOpus 5.5 is so unbelievably good that people are starting to say that you can genuinely build most SaaS apps in a few hours.16.6x
346K@kanekallawayMeta just had a crazy product launch #ai #tech #meta #muse [caption]10.8x
114K@taki.gptClaude, edit this video, add these effects, orbit, chat takeover, frame escape, snap zoom, checklist pop, headline storm, zoom portal, growth curve, studio reveal, b-r...38.3x
62K@nomadatoastI'll show you why you need to be really careful with the eye right now.14.6x
56K@fawziammacheAnthropics IPO financials have been leaked and here are some of the most shocking numbers that you'll see.11.5x
50K@tony.aubeThere’s a really cool consumer AI startup hiding behind this! [caption]

Full data: creators.csv · posts.csv

Instagram, your direction

Copy what works, bar for bar

Free, useful content the way Dhruv does it, with your face on an AI twin, in English, sounding technical. We took apart 73 face-on reels second by second (Dhruv's 15 plus the best and a typical reel from 18 face-on creators): every cut, caption, beat and opening line. The structure is copied exactly. The words are new, because Instagram stops recommending accounts that post copies, and copied scripts get called out.

What Dhruv's two winners did

@maujawaan.tech (end card: "Dhruv Mahajan | Tech & AI"). 15 reels, usual reel about 2K plays.

  • 1.68M: a free tool that finds hidden cameras on an Airbnb's wifi. 467K: a mock-interview tool posted on Dussehra, tied to "your comeback".
  • Big stakes everyone feels (family safety, a new job). His Apple-trivia reels with the same polish got 1.5K.
  • The payoff is on screen by 0.7 seconds: a dated news card or a festival calendar card in the top third.
  • Same script posted 3 times: the version opening on a giant title got 63.5K; the two opening on "a friend of mine..." got 1.5K and 1.2K.
  • Short: 23 and 41 seconds, spoken fast. A comment keyword at the end ("comment LANKA") plus "send me this reel in DM", which turns the ask into a share: 11,656 comments.

What 42 face-on winners have in common

  • One high-stakes sentence from the first frame, often with a number or "just / can now" (opusjake: "Claude just killed day traders", 1.55M).
  • Text or graphics already on screen at 0.7 seconds: 38 of 42 winners.
  • One named, free, public tool: "It's called X and it's free."
  • Face in the bottom half, cards changing on top. A face never holds more than 8 seconds (median shot 3.3 s vs 8 s in typical reels), which suits an AI twin.
  • 24 to 63 seconds, about 220 words a minute.
  • Comment keyword in the last 3-6 seconds and again in line one of the caption. Same video, same day: 242K with the caption keyword, 3K without.
  • Repost every script 2-3 times as Trial Reels with a different first frame. Sabrina's identical repost got 154K vs 1.8M.

12 templates to copy (exact beat timing: templates.json, all 73 reels: reels.csv)

T01 · 42 s · free-value tip (Dhruv style)

Fear check: dated headline + free tool + 'if you find one' checklist

Fully automatable

826x his median: a mass-audience safety fear, a dated third-party headline as the hook, a free public tool and a 3-step checklist. The same polished format on niche Apple trivia got ~1.5K (DetuCilTtKz, DepfkwKz9Nl).

Copied from: maujawaan.tech 1.68M

T02 · 24 s · AI-employee proof or free-value tip

Calendar-moment hook + insider proof + pun keyword + 'DM me this reel'

Fully automatable

230x his median. Same product as his 7,480-play MOCK reel; the winner is 24 s not 57 s, opens on the calendar moment, uses proof (a friend's offer) instead of a coupon, and was posted on the day of the moment (2026-10-20, tagged #dussehra2026).

Copied from: maujawaan.tech 468K · maujawaan.tech 7.48K

T03 · 35 s · free-value tip

Split-screen free skill: 'It's called X and it's completely free'

Fully automatable

opusjake's two winners (128x, 92x) and kayvon's ECC (124x) share it: one free named tool, a vivid or high-stakes first line, a 2-step install, keyword DM. opusjake's typical reel (0.98x) dropped the face and the spoken hook.

Copied from: opusjake 2.16M · opusjake 1.55M · kayvon.ai 1.6M

T04 · 58.5 s · free-value tip

Numbered codes list under a persistent title

Needs a real screen recording

60x median. Visual output in every item (images, not text walls); a list keeps people watching to the next number; save-worthy. The identical repost got 154K, so even a repost of a proven list does 5x.

Copied from: sabrina_ramonov 1.8M · sabrina_ramonov 154K

T05 · 25-39 s · AI-employee proof

Agent-factory tour: 'I'm the only human in my company'

Needs a real screen recording

319x and 488x median on a small account. The footage is his own working system, not a demo of someone else's tool. His typical reel had no pill, no captions and only 11 s of footage.

Copied from: jakedavidsai 528K · jakedavidsai 808K

T06 · 35-46 s · free-value tip

Open-source repo that does a business job, with a live result counter

Needs a real screen recording

liam 45x: one money outcome shown in frame 0 (rows counting), named open-source tool with stars, cost comparison. His typical reels were abstract ('9 techniques'). Jack 11x: free repo + star count vs his typical sponsored tool pitch.

Copied from: liamjohnston.ai 250K · jackroberts___ 95K

T07 · 52 s · AI-employee proof + tip

'This is how to turn Claude into X': hidden fact → why it matters → my routine → favourite part → install

Fully automatable

75x median: one promise that applies to every Claude user, proven with a real file path on screen. His typical reel was a 75 s listicle of niche design tricks.

Copied from: austin.marchese 146K · austin.marchese 26K

T08 · 59 s · free-value tip

'Apparently [they] are reverse-engineering [X]': copy-the-system story with the keyword in the caption

Fully automatable

36x median; reverse engineering is the hottest topic in this sample (varun 947K, trakin 474K, angus 242K, nick 123K).

Copied from: angus.sewell 242K · angus.sewell 3.17K · trakin.ai 474K

T09 · 63-80 s · free-value tip (news)

News split-screen: 'This repo is quietly blowing up' / 'Breaking news'

Fully automatable

2.35x on a 403K median: one clean news hook with numbers in 80 s vs his 136 s staged office skit (0.98x).

Copied from: thevarunmayya 947K · thevarunmayya 435K

T10 · 108 s · AI-employee proof

Live demo proof: 'I gave my AI [a phone]'

Needs a real screen recording

7.7x: a live, novel demo with a time-boxed promise; his typical reel was a news recap.

Copied from: realrileybrown 442K

T11 · 42.5 s · free-value tip

Bad / Good / Great row

Fully automatable

2x her median: the payoff graphic is visible in frame 0 and the format repeats; her typical reel was an 80 s case study.

Copied from: personalbrandlaunch 70K

T12 · 79.5 s · free-value tip (trust)

Debunk the viral AI claim

Needs you on camera

13.6x: a profanity-level reaction + the source clip in frame 0. Her typical debunk (14.5K) had a calmer hook and an obscure subject.

Copied from: jiannacapri 170K

15 scripts ready to produce, in English

11 are free-value tips (Dhruv's lane), 4 are proof from your own AI employees. Anything in [brackets] is a real number you supply. Each lists what must be checked before it posts.

S01 · Free Claude skill that checks a rental's wifi for hidden cameras (same topic as source) · comment CAMERA · 38.6 s · free-value tip (Dhruv style)

Template T01, copied bar for bar from maujawaan.tech, 1.68M plays. Caption: Comment CAMERA and I'll send you the skill 👇 Check the wifi before your family unpacks. #airbnb #privacy #claude #ai

0.0-4.5
In March 2024, Airbnb banned indoor security cameras in every listing. Banned does not mean gone.
On screen: News card: 'Airbnb bans indoor security cameras' + publisher + date. At ~3.8 s giant red title: BANNED. NOT GONE.
4.5-9.2
There's a free, open-source Claude skill called privacy-scanner that checks the room's network for you.
On screen: Repo card: privacy-scanner · Free · open source · Claude skill
9.2-13.1
Join the rental's wifi, then tell Claude to run it. That's the whole setup.
On screen: Card: Rental WiFi ✓ connected / chat: 'run privacy-scanner'
13.1-20.9
It lists every device on that network, looks up who made each one, flags known camera brands, and checks which device is sending video out of the building.
On screen: Devices on this wifi: [n] / Camera maker found: [n] / Sending video out: [n] (numbers from a real test run)
20.9-23.4
You get one report, ranked from safe to critical.
On screen: Report card with one red CRITICAL row
23.4-28.4
If something shows up critical, don't unplug it. Photograph it where it sits, and report it to Airbnb.
On screen: IF YOU FIND A CAMERA: 1 Don't unplug or touch it 2 Photo of the device and where it is 3 Report it to Airbnb
28.4-33.7
It only sees devices on that wifi, so still check anything aimed at the bed. Keep your family safe.
On screen: First night: ✓ Join the wifi ✓ Run the scan ☐ Check chargers and clocks
33.7-36.8
Follow me, and comment CAMERA, and I'll send you the skill.
On screen: Comment box typing CAMERA, then giant gold CAMERA
36.8-38.6
On screen: @[handle] end card

Check before posting: privacy-scanner exists, is free, and works: run it on our own network with a known camera; the skill only does passive device discovery (no attacks on other devices); Airbnb's March 2024 indoor-camera ban: link the source on the card

S02 · AI receptionist for the holiday rush. Source tool is unnamed on screen, so the topic chang · comment DIYA (non-Indian version: RUSH, moment = Black Friday to Christmas) · 25.6 s · AI-employee proof

Template T02, copied bar for bar from maujawaan.tech, 468K plays. Caption: Light up every call this season 🪔 Comment DIYA or DM me this reel. #diwali2026 #smallbusiness #ai

0.0-7.0
If this reached you between Diwali and New Year's, your phone is about to ring more than any other week. Don't let it ring out.
On screen: Timeline card: HOLIDAY SEASON · DIWALI ●-----● NEW YEAR'S, dot sliding right; 'Diwali' in gold serif
7.0-12.0
And if the one answering already knew your prices, your hours and your calendar, what's better than that?
On screen: Call card: AI RECEPTIONIST · LIVE / title WHAT'S BETTER?
12.0-18.4
That's an AI receptionist. I set one up for [client type], and it now picks up [real number] after-hours calls a week.
On screen: Real call-log card (client name hidden)
18.4-22.6
Want the setup? Comment DIYA, or DM me this reel, and I'll send it over.
On screen: Comment box typing DIYA + share sheet 'Send to @[handle]'
22.6-24.3
Light up every call this season.
On screen: LIGHT UP EVERY CALL + giant DIYA on warm diya-light background
24.3-25.6
On screen: end card + profile pill

Check before posting: the call count is from the real log; the client agreed to be shown, even anonymised (bead trejon-pogp)

S03 · Arena skill: many Claude sub-agents compete on one task (same topic as source) · comment ARENA · 40.8 s · free-value tip (Dhruv style)

Template T03, copied bar for bar from opusjake, 2.16M plays. Caption: Comment ARENA and I'll send you the link 👇 A hundred Claudes, one task, only the best draft survives. #claude #claudecode #ai

0.0-4.7
When Claude hands you safe, generic answers, stop re-prompting. Make a hundred copies of it compete.
On screen: pills: SAFE, GENERIC ANSWERS / A HUNDRED COPIES
4.7-7.2
The skill is called Arena, and it costs nothing.
On screen: name card: Arena · Claude skill
7.2-11.4
Use it when good enough isn't: a cold email, a pricing page, a product name.
On screen: pills
11.4-21.2
Arena hands the same task to a hundred sub-agents. Each one takes a different angle: shorter, more data, the opposite opinion. Then they critique each other and cut the weakest drafts, round after round.
On screen: x100 (huge)
21.2-25.1
The part that matters: the survivor has already been attacked by ninety-nine critics.
On screen: 1 LEFT
25.1-29.6
So you stop rolling the dice ten times and get the strongest version in one run.
On screen: before/after
29.6-33.8
Setup is one step: paste the skill link into Claude and say, install this skill.
On screen: link → 'install this skill'
33.8-37.7
Fair warning: a hundred agents burns tokens, so save it for work that counts.
On screen: TOKENS
37.7-40.8
Follow me and comment ARENA, and I'll send you the link.
On screen: 1 Follow 2 Comment ARENA

Check before posting: the Arena skill exists, its link, and that it really runs ~100 sub-agents with critique rounds; run it once and note the real token cost

S04 · Free LLM API: keeps Claude Code running on free models after you hit your limit (same topi · comment CLAUDE · 34.6 s · free-value tip (Dhruv style)

Template T03, copied bar for bar from nick_saraev, 467K plays. Caption: Comment CLAUDE and I'll send you the link 👇 Keep Claude Code going after your limit, on free models. #claudecode #ai #opensource

0.0-4.2
You can keep Claude Code running after you hit your limit, and it costs nothing.
On screen: pills: AFTER YOUR LIMIT / COSTS NOTHING
4.2-9.5
Someone on GitHub built a router that pools the free tiers of dozens of AI providers into one connection.
On screen: [N] free tokens a month
9.5-13.4
That means models like Gemini, Kimi, Qwen, GLM and Mistral, all behind one key.
On screen: logos in sequence
13.4-17.6
The project is Free LLM API, and one setup step connects it to Claude Code.
On screen: Free LLM API
17.6-22.6
When you hit Claude's cap, it moves your session to the next best free model instead of stopping.
On screen: CAP HIT → SWITCHED
22.6-27.1
And it writes a handoff note first, so the new model knows what you were doing.
On screen: HANDOFF NOTE
27.1-32.4
Two catches: free models are weaker, and your code goes to whoever hosts them. Keep client work off it.
On screen: 2 CATCHES
32.4-34.6
Comment CLAUDE and I'll send you the link.
On screen: Comment CLAUDE

Check before posting: Free LLM API exists; its real model list and free-token figure; the handoff-note behaviour is real (test it)

S05 · 10 prompt shortcuts that make an AI teach you anything (same topic; our own set of codes, · comment CODES · 48.6 s · free-value tip (Dhruv style)

Template T04, copied bar for bar from sabrina_ramonov, 1.8M plays. Caption: Comment CODES and I'll send you 50 more 👇 Save this for the next thing you need to learn. #chatgpt #claude #ai #learning

0.0-2.2
Ten codes that make AI teach you anything.
On screen: title box: Secret Codes to Master Any Topic 🤯
2.2-7.2
Number one: ELI5. Add it to any question and you get the version a curious kid could follow.
On screen: #1 ELI5
7.2-12.2
Two: 80/20. It gives you the twenty percent of the topic that does eighty percent of the work.
On screen: #2 80/20
12.2-16.4
Three: ANALOGY, then your job. It explains the idea with examples from your own industry.
On screen: #3 ANALOGY: [your job]
16.4-20.0
Four: MIND MAP. You get a visual map of how every idea connects.
On screen: #4 MIND MAP
20.0-25.0
Five: QUIZ ME. One question at a time, and it won't move on until you get it right.
On screen: #5 QUIZ ME
25.0-29.2
Six: TEACH BACK. You explain it in your words, and it grades what you missed.
On screen: #6 TEACH BACK
29.2-33.1
Seven: COMPARE TABLE. Two options side by side, in a table you can screenshot.
On screen: #7 COMPARE TABLE
33.1-37.0
Eight: FIRST PRINCIPLES. It strips the topic to the basic facts and rebuilds it.
On screen: #8 FIRST PRINCIPLES
37.0-40.3
Nine: COMMON MISTAKES. The five errors beginners make, before you make them.
On screen: #9 COMMON MISTAKES
40.3-43.6
Ten: CHEAT SHEET. The whole topic on one page you can print.
On screen: #10 CHEAT SHEET
43.6-48.6
They're not hidden commands, just shorthand the model already understands. Follow me and comment CODES for fifty more.
On screen: Follow me / comment 'CODES' / ♻️ share for priority

Check before posting: record each output for real in ChatGPT or Claude; pick prompts whose output is visual; the 50-code list exists before posting (it is the DM)

S06 · Tour of his own SAGE agent fleet on the 3D roster dashboard (his proof; the format's topic · comment TEAM · 31.2 s · AI-employee proof

Template T05, copied bar for bar from jakedavidsai, 528K plays. Caption: My company runs on AI employees. Comment TEAM and I'll show you how to build your first one. #aiagents #ai #smallbusiness

0.0-4.5
I run my company with [N] AI employees, and every one of them has a desk.
On screen: persistent pill
4.5-6.0
This is the office.
On screen: pill
6.0-9.6
Trejon, my chief of staff. Plans the day, writes the code, remembers everything.
On screen: label: Trejon
9.6-13.5
Coach. Sends my eight a.m. scorecard and pings me when an agent gets stuck.
On screen: label: Coach
13.5-16.6
The note-taker. Joins my client calls and files the transcript.
On screen: label: [dashboard name]
16.6-19.7
Envoy. Writes and sends the outbound emails: [real number] this week.
On screen: label: Envoy
19.7-21.7
Victor. Runs [real task] for Opti Plaza.
On screen: label: Victor
21.7-25.9
And every morning at four-thirty, they review the day and update their own notes.
On screen: 4:30 AM
25.9-27.9
If one gets stuck, Coach tells me.
On screen: pill
27.9-31.2
Comment TEAM and I'll show you how to build your first one.
On screen: Comment 'TEAM' for the setup!

Check before posting: dashboard labels match the real roster; every number is from a real log; which agents and clients may be shown (bead trejon-pogp); sage-1 is healthy before recording (overloaded on 2026-10-21, bead trejon-fbvgs)

S07 · Claude + ScrapeGraphAI builds free local-business lead lists (same topic as source) · comment LEADS · 45.4 s · free-value tip (Dhruv style)

Template T06, copied bar for bar from liamjohnston.ai, 250K plays. Caption: Comment LEADS and I'll send you the repo and my exact prompt 👇 Public data only. #claude #leadgeneration #smallbusiness

0.0-3.6
Claude can now build you a list of local business leads, for free.
On screen: pills: LOCAL LEADS / FOR FREE
3.6-7.8
Under the hood is ScrapeGraphAI, an open-source scraper with about [N] stars on GitHub.
On screen: star counter
7.8-13.9
Claude uses it to search the web or walk through directories, and pulls the public details you need: company, owner, phone, website.
On screen: column headers appear
13.9-19.8
Everything lands in one tidy spreadsheet. It all happens on your own machine, so no scraping service bills you per row.
On screen: $0 vs [paid service price]
19.8-24.8
Setup is one message: drop the GitHub link into Claude and ask it to set the tool up.
On screen: GitHub link → 'set this up'
24.8-29.8
Here's a live run. I ask for [dentists in Surrey with no online booking], and rows start appearing.
On screen: the request text
29.8-35.1
Save it as a workflow, schedule it for each morning, and new leads are waiting when you sit down.
On screen: 7:00 AM Mon-Fri
35.1-39.3
One rule: public data only, and check Canada's anti-spam law before you email anyone.
On screen: CASL
39.3-45.4
Comment LEADS for the repo and my exact prompt, and forward this to whoever on your team still builds lists by hand.
On screen: LEADS (giant)

Check before posting: install and run ScrapeGraphAI once; record the real counter; star count; the paid-service price used in the comparison

S08 · MiroFish: thousands of simulated customers react to your launch (same topic as source) · comment CROWD · 30.9 s · free-value tip (Dhruv style)

Template T06, copied bar for bar from jackroberts___, 95K plays. Caption: Comment CROWD and I'll send you the link 👇 Test the launch before you pay for ads. #opensource #ai #startup

0.0-6.1
A free, open-source tool now puts your launch in front of thousands of AI customers before a real one sees it.
On screen: pills
6.1-9.4
The project is MiroFish, and [N] people have starred it on GitHub.
On screen: Star [N]
9.4-12.5
Feed it your landing page, your pricing, or an ad draft.
On screen: 3 input cards
12.5-15.6
Then ask one question: will people want this, or scroll past?
On screen: prompt box
15.6-21.7
This isn't one chatbot's opinion. MiroFish creates a crowd of customers with different jobs, budgets and moods, and each one decides alone.
On screen:
21.7-27.0
Treat it as a smoke test, not a survey: it catches the obvious misses before you spend on ads.
On screen: SMOKE TEST
27.0-30.9
Want to test your next launch? Comment CROWD and I'll send you the link.
On screen: "crowd"

Check before posting: MiroFish exists, star count, and it does run multi-persona simulations (run it once on a real page)

S09 · Claude Code keeps every session as a local file; his chief of staff reads them nightly to · comment IMPROVE · 48.0 s · AI-employee proof + tip

Template T07, copied bar for bar from austin.marchese, 146K plays. Caption: Comment IMPROVE and I'll send you the prompt and the folder paths 👇 #claudecode #aiagents #ai

0.0-4.2
This is how to make Claude Code get better at your work every single night.
On screen: serif title: Self-Improving Claude
4.2-12.6
Every session you run is saved, and not just on Anthropic's side. It's a file on your own computer, in a hidden folder called dot claude, projects. Here's the path.
On screen: ~/.claude/projects/<project>/<session>.jsonl
12.6-19.0
Why does that matter? That folder records every time it got something wrong and you corrected it. So why is nobody reading it?
On screen: Gold mine of corrections
19.0-26.5
My AI chief of staff reads the day's sessions at four-thirty every morning. It updates its notes, its rules and its memory from what went wrong.
On screen: 4:30 AM · memory-consolidate
26.5-33.8
Here's my favourite part: before it saves any change, it runs its own tests. If a test fails, it rolls the change back and messages me.
On screen: TESTS → ROLLBACK → ALERT
33.8-38.0
So I'm not writing rules by hand. It learns from the work it already did.
On screen:
38.0-44.7
To set it up, make one skill that reads yesterday's files, lists the corrections, and proposes three rule changes. Run it on a schedule.
On screen: 1 read 2 list 3 propose
44.7-48.0
Comment IMPROVE and I'll send you the prompt and the folder paths.
On screen: Comment "IMPROVE"

Check before posting: the session path on his Mac matches the card; the 04:30 job and its rollback/alert behaviour as described in AGENTS.md (memory-consolidate)

S10 · Reverse-engineer an expensive coaching funnel from public pages with Claude sub-agents (sa · comment FUNNEL · 45.8 s · free-value tip (Dhruv style)

Template T08, copied bar for bar from angus.sewell, 242K plays. Caption: If you want the prompt, comment FUNNEL 👇 (keyword in line one: the same reel without it got 3K plays, with it 242K)

0.0-4.5
Apparently, people are using Claude to take apart the funnels behind ten-thousand-dollar coaching programs.
On screen: pill: How to reverse-engineer a $10k coaching funnel / 2-line caption
4.5-9.8
Open any big coach's testimonial page. Every client says they went from stuck to fully booked right after joining.
On screen: 2-line caption
9.8-15.1
So the program is mostly the sales system. Wait, that means you could study that system from the outside.
On screen: 2-line caption
15.1-21.5
Screenshot the testimonials. Have Claude find each client's public business page. Then pull their ads from Meta's Ad Library, which anyone can search.
On screen: Meta Ad Library
21.5-36.0
Now spin up sub-agents with one prompt: visit each page as a normal visitor, sign up for the free lead magnet with your real email, map every step and message, and list what repeats. The patterns that show up in every funnel are the system you were going to pay for.
On screen: prompt text
36.0-40.2
Here's what Claude built me: a quiz funnel and a set of follow-up messages.
On screen:
40.2-45.8
You don't need a ten-thousand-dollar program to see the playbook. Follow me, I'll keep posting these for free.
On screen:

Check before posting: run it on a real public funnel and keep the outputs; no personal data of testimonial clients shown; blur names and faces

S11 · REA (Reverse Engineer Anything): an MCP that connects coding agents to Ghidra (same news i · comment none (question CTA, like the source) · 52.3 s · free-value tip (Dhruv style) (news)

Template T09, copied bar for bar from thevarunmayya, 947K plays. Caption: REA is free and it's trending. Which app would you take apart first? #opensource #claudecode #ai

0.0-6.7
This GitHub repo is climbing fast because it lets an AI agent explain how any app works, even when you don't have its code.
On screen: word-by-word caption
6.7-12.3
The project is REA, short for Reverse Engineer Anything. No cost, it runs locally, and it sits at [N] stars.
On screen:
12.3-29.9
Here's the problem it solves. When software ships, the readable code is gone, and what's left is machine instructions nobody wants to read. REA plugs agents like Cursor, Codex and Claude Code into Ghidra, the free reverse-engineering suite the NSA released. You point it at a program and ask, in plain English, what a feature does. It traces the binary and explains.
On screen:
29.9-37.2
When people on Reddit were asked what they'd use it on, the answers were old game files, smart-TV firmware, and apps nobody can update anymore.
On screen: highlighted thread lines
37.2-42.5
One developer says he pulled a Steam game apart and had it running in Unity in about thirty minutes.
On screen: highlighted post
42.5-49.5
The bigger shift: software you own stops being a black box. One rule: only do this on software you own or have permission to inspect.
On screen:
49.5-52.3
Which app would you take apart first? Tell me below.
On screen: caption

Check before posting: REA repo, name, star count, and that it is still trending on the post date; the Reddit thread and the Steam/Unity post exist (link them); Ghidra NSA release (2019) is correct

S12 · His AI receptionist answering a real after-hours test call (format same; the capability ch · comment PHONE · 55.7 s · AI-employee proof

Template T10, copied bar for bar from realrileybrown, 442K plays. Caption: Comment PHONE and I'll send you the FAQ template and hand-off rules 👇 #aireceptionist #smallbusiness #ai

0.0-4.5
You can give an AI employee its own phone number in one afternoon. Here's mine answering.
On screen: hook line
4.5-9.0
It's [time] on a [day], the office is closed, and I'm calling my own business line.
On screen:
9.0-35.0
(real call plays: the agent answers, gives hours and prices from the FAQ, then books or takes a message)
On screen: live call captions
35.0-39.2
I hung up, and [what really happened next: summary text, calendar entry] arrived [how fast].
On screen:
39.2-51.2
What you need: a phone number from [provider], a voice agent, and a one-page FAQ with your prices, your hours and the three questions you get most. The hard part is the rule for when it hands the call to a human.
On screen: 1 number 2 voice agent 3 one-page FAQ 4 hand-off rule
51.2-55.7
Comment PHONE and I'll send you the FAQ template and the hand-off rules I use.
On screen: comment PHONE

Check before posting: record a real call; claim only what the recording shows; name the real stack

S13 · Bad / good / great AI employee setup. DEVIATION: the source topic is Instagram metrics, wh · comment none (follow CTA, like the source); optional JOB for the Starter Kit job descriptions · 46.3 s · free-value tip (Dhruv style)

Template T11, copied bar for bar from personalbrandlaunch, 70K plays. Caption: Bad, good and great AI employees. Which column is yours? #aiagents #smallbusiness #ai

0.0-4.5
This is a bad AI employee. This is a good one. This is a great one.
On screen: BAD GOOD GREAT (active word in lime)
4.5-11.2
Job description: bad is 'be helpful'. Good is 'answer pricing questions'. Great is 'answer from this price sheet, and text me if you're unsure'.
On screen: JOB DESCRIPTION
11.2-16.8
Data: bad is the open internet. Good is your website. Great is your own price list, calendar and past emails.
On screen: DATA
16.8-24.1
Hand-off: bad is never. Good is 'when asked'. Great is a clear rule, like every refund or complaint goes to a person within two minutes.
On screen: HAND-OFF
24.1-28.8
Memory: bad forgets every chat. Good remembers the customer. Great remembers the customer and what they bought.
On screen: MEMORY
28.8-34.4
Testing: bad is launch and hope. Good is ten test chats. Great is fifty real past questions, scored before launch.
On screen: TESTING
34.4-39.1
Cost: bad is unknown. Good is a monthly bill. Great is cost per conversation, tracked every week.
On screen: COST
39.1-43.8
Owner: bad is nobody. Good is IT. Great is one named person who reads the weekly log.
On screen: OWNER
43.8-46.3
Follow for more on AI employees that actually work.
On screen: FOLLOW FOR MORE

Check before posting:

S14 · Debunk the recycled 'my AI agent runs my business while I sleep' videos (same topic: recyc · comment FLAGS · 40.1 s · free-value tip (Dhruv style) (trust)

Template T12, copied bar for bar from jiannacapri, 170K plays. Caption: Comment FLAGS and I'll send you my checklist for judging any AI claim 👇 #ai #aiagents #scam

0.0-6.4
This video is everywhere right now: my AI agent runs my business while I sleep. Most of them are the same script, recycled.
On screen: pill: Biggest AI agent claim 🚩🚩
6.4-16.2
I build AI employees for real businesses, so I know what a working one looks like. It has logs, it has costs, and it has a person who checks it. Here's what to look for.
On screen: captions
16.2-26.5
First red flag: no screen recording of the agent actually running, just a dashboard screenshot with no date. Second: the same hook, word for word, across ten accounts. Third: no skills needed, followed by a paid course.
On screen: 1 NO LIVE RUN 2 SAME HOOK 3 PAID COURSE
26.5-36.5
Here's what's true. Real agents break. They need a clear job, access to your data, and a rule for when to hand off to a human. Anyone who shows you that part is worth listening to.
On screen:
36.5-40.1
Comment FLAGS and I'll send you my checklist for judging any AI claim.
On screen: Comment FLAGS

Check before posting: do not name or show any private individual; blur every handle

S15 · ECC (Everything Claude Code): one developer's free plugin with dozens of agents and skills · comment ECC · 38.4 s · free-value tip (Dhruv style)

Template T03, copied bar for bar from kayvon.ai, 1.6M plays. Caption: Comment ECC and I'll send you the link and my install notes 👇 #claudecode #opensource #ai

0.0-3.9
One developer just gave away a full engineering team for Claude Code, for free.
On screen: script-italic: One developer / FREE
3.9-10.6
[63] agents, [249] skills, and a security agent that attacks your own code before anyone else does. The project is ECC, free on GitHub.
On screen: 63 AI AGENTS / 249 SKILLS / FREE ON GITHUB
10.6-14.5
One person built it in about ten months, and it has passed [240,000] stars.
On screen: [240,000] STARS
14.5-19.5
Install is two commands in Claude Code: add the repo to the plugin marketplace, then install the plugin.
On screen: /plugin marketplace add [repo] → /plugin install [name]
19.5-30.9
After that, you give orders. Plan this feature gets you an architect's design. Build it produces the code. Secure it sends an agent to attack your app and patch the holes. Test it keeps tests running until coverage reaches eighty percent.
On screen: PLAN / BUILD / SECURE / TEST
30.9-35.9
Before you install any plugin, read what it's allowed to run. Skills can execute code on your machine.
On screen: READ FIRST
35.9-38.4
Comment ECC for the link and my install notes.
On screen: COMMENT "ECC"

Check before posting: repo name, agent/skill counts, star count and build time; the two install commands work as written; the 80% coverage behaviour is real

What the twin cannot do well: walking shots, big hand gestures, long emotional takes, props held to the lens. So we skip those formats (Aryan's mansion walk, Jun Yuh's life b-roll, Gianna's debunks) and let fast cards carry the energy. Proof reels (your agent dashboard, a real test call) still need one real screen recording each, which we make ourselves.
What to post

50 post ideas, each copied from a real winner's structure

The structure comes from a post that already won (linked). The substance is your own work. Anything in [brackets] is a real number or name you supply or approve; nothing is invented.

Download: topic-bank.csv

How it runs

No daily routine for you

Your part

  1. Once, about 90 minutes. OK the position, give the standing go, pick which client results can be public, confirm handles. Send 5-10 photos and a voice sample once so the AI twin can be built.
  2. Weekly, 15-20 minutes on your phone. Approve the batch: OK, Edit or Kill. Skip a week and nothing posts, nothing breaks.
  3. Whenever you feel like it. A voice note with a take or a story, or a selfie clip. Your dictation already flows in.

Trejon's part

  • Weekly radar of each platform's breakout posts
  • Drafts from real work: closed jobs, agent logs, client results you approved, Cherry Shot outputs
  • Visuals: reels from screen recordings with captions, cards, carousels, screenshots
  • Quality check, scheduling after your OK, first-hour reply drafts
  • Weekly numbers: views vs your median, saves, keyword comments, owner DMs, calls
Why the standing go matters. The content engine has written drafts every morning since 26 September, 26 days in a row, and posted nothing. The blocker is approving and posting, not ideas.
From viewer to client

One free kit, two paths

Post→Comment KEYWORD (Instagram) or link in first reply (X, LinkedIn)→Free AI Employee Starter Kit on gurleenkhurana.com→Owners: 20-minute AI employee audit → SAGE installorBuilders: email list

The kit: job descriptions for 5 common AI employees, the 4-question scope test, a cost sheet. Market range for an install (our research, not a price list): $750 to $3,500 setup plus $300 to $1,200 a month.

No community at launch. Under 10K followers with a real skill, the evidence points to a high-value service, not a $49 community (Jun Yuh, Dan Koe). A community needs daily presence: Justin Welsh shut a $15K-a-month one because he “had to be on Slack 24/7”.

First 30 days

Launch plan

Days 1-3: setup

Your one-time items. Trejon points the content engine at your brand, builds the phone approval page, sets up scheduling, writes the Starter Kit and the gurleenkhurana.com page with a booking link, and makes reel templates that need no filming.

Week 1: the flagship

“My company runs on AI employees” everywhere: X Article (idea 2), proof reel (idea 1), LinkedIn post (idea 4), TikTok cross-post. Plus 2 reels on this week's AI news in business terms.

Week 2

Installs pillar, the first free-resource keyword reel (idea 37), one honest-numbers post.

Weeks 3-4

Remake whatever broke out, in other formats. First YouTube video if you record the 30-minute tour (idea 10). Day-30 review: which pillars to weight up on each platform.

Numbers to judge by (targets, not forecasts)

WhenTargetDepends on
Day 30One post per platform at 5x your median; 10+ owner DMs or keyword requests; first audit callThe standing go, so posts actually go out
Day 90Instagram 2-10K followers; LinkedIn +1-3K; X 300-1,500; 5-15 audit calls; 1-2 SAGE installs5 reels + 4 LinkedIn posts + 1 Article every week, and at least two reels passing 100K. Estimates.
TodayX @gurleenAI: 5 followers, best post 1,111 viewsStarting point

Costs: Everything except the AI twin runs on tools we already have. Instagram AI twin on Higgsfield: about $3.30 per reel as a draft and $7.70 as a final (5 reels a week is roughly $165 a month). Later, each with your go: comment-to-DM automation, about $15 a month; X Premium if Articles need it, about $8 a month (not yet verified).

Blocked on you

Five decisions

  1. The position: “I run my company on AI employees, and I build them for real businesses. I show the receipts.” Yes, or edit it.
  2. Standing go: posts you approved go out on your accounts at scheduled times, and reply drafts you approved post too. Without it, nothing posts.
  3. Public proof: which clients and results can be shown by name, and which only anonymised (“a Vancouver clinic”).
  4. One flag on your public-copy rule. “Only successes, never name AI tools” was set for company sales pages. On personal channels the winning formats name the tools (“it's called X and it's free”) and show real costs and fixes. Proposal: your personal channels may do both, framed as what got fixed and never blaming a client; company pages stay as they are. Your call.
  5. Your face and voice: decided. An AI twin on Higgsfield, English only. To start it I need 5-10 clear photos of your face, a voice source (old YouTube voice-overs or a 2-minute voice note), and Higgsfield credits: the account has 0.85 left. About $3.30 per reel as a 480p draft, $7.70 as a 720p final. Also: which Instagram account to use.
The people you named

Who they are and what to take from them

NameWhereWhat to take
Nick Saraev@nick_saraev (IG), YouTube 566KFree tools reels with a comment keyword; Maker School $184/mo
Nate HerkYouTube 1.09M, free Skool 474KYouTube-led; paid community $129/mo, 3.4K members; no reels in window
Liam OttleyYouTube 862K, IG @liamottley 44KBusiness-rebuild videos; agency and accelerator
Brendan Jowett ("Brandon Chowett")YouTube @brendanautomation 63KOne 824K launch-week video; paid community has 115 members
Rowan CheungX 601K, IG 502KAI news; Rundown newsletter 2M+, ad-funded
Marc LouX @marclou 415KBuilds in public; products $94K/mo self-reported
levelsioX 980KShips products, posts revenue; Photo AI $86K/mo self-reported
Varun MayyaIG @thevarunmayya 1.2M, YouTube 1.18MIndia's top AI news explainer; agencies and edit school
Jun YuhIG @jun_yuh 5.3MPersonal-brand teacher; the 3-1-2-1 weekly mix
aryan.mjnIG Aryan Mahajan 36K5.23% comment rate on a keyword reel
Alex HormoziIG 5M, YouTube 4.58MProof first; his top-viewed videos made zero sales
Personal Brand LaunchIG @personalbrandlaunch (Ava Yuergens) 1MStudy small creators who blew up; double down on 5x outliers
Gianna CapriIG @jiannacapri 216KPublic "$100K in 90 days with AI" challenge series, free Skool
Creator CampIG @creatorcamp 313KFilm and creator events, not AI; 2 reels in window
Dhruv MahajanLikely IG @maujawaan.tech (65%, please confirm)Hinglish tech reels; 1.68M reel on about 20K followers
"Angus Dartsville"Likely IG @angus.sewell, Angus Sewell McCann (70%)Informative AI how-tos for non-technical people, 195K
Robert Macchiavelli, World of MayaNot foundSend the handles and they will be added
Superpowered.xyzA parked domain for saleNo account behind it
Sources and limits

Where this comes from

Pulled 21 October 2026, window 23 July to 21 October. YouTube Data API (110 channels, 7,497 uploads); X via a read-only session (41 accounts, 2,385 posts); Apify for Instagram (50 accounts, 897 reels), TikTok (31 profiles) and LinkedIn (37 profiles, 580 posts), $3.54 in total; Pew Research 2026; X's open-source ranking code; LinkedIn and YouTube official posts; public Skool pages; 1,982 YouTube and X comments; transcripts of the top posts.

Not confirmed: Instagram's AI-label rule; whether X Articles need Premium; LinkedIn's link penalty (third-party only); most revenue figures (self-reported, marked in the data); the Angus and Dhruv matches. Some samples cover the newest 10-36 posts per account rather than the full 90 days.

Data: topic bank · Instagram · TikTok · YouTube · X · LinkedIn