TikTok outliers · this fortnight
AI founders
15 accounts · 0 copies · last scan 8h ago
Founders and builders posting about AI tools, vibe coding and building in public. The landing-page demo niche.
#aifounder#buildinpublic
Free to copy. Scanning uses your plan.
What is working in AI founders right now
Computed from the outliers below that were analyzed from the actual video or slides: formats, hook techniques, mechanisms, length, topics, then a written read with hooks to try and next post ideas.
Talking-head videos with bold-claim hooks about specific AI tools (Claude, Grok, Codex) are dominating, with novelty/contrarian angles far outperforming generic utility tips.
9 outliers in the last 14 days, 8 analyzed from the video · median 6.2× · median length 60s · speech 100% · music 38% · captions 100% · 51d ago
Talking-head is the workhorse format (63% of outliers, mean 20.5x) and bold_claim hooks show up in 75% of winners at 20.1x mean. The single biggest hit, @20vc_tok's Dyson toothbrush reaction, shows novelty/absurd-gadget framing can spike 72x even outside core AI tooling. Contrarian tool-rivalry takes (@sabrina_ramonov on Claude vs ChatGPT) and specific-number claims (@ai.with.andrew's $10k Claude Fable websites) both cleared 18x+, while plain utility explainers (@aistartupfren, @nate.b.jones) barely beat baseline at ~1x. No-CTA posts actually out-multiplied comment-CTA posts (24.1x vs 9.9x mean), suggesting the hook and topic are doing the heavy lifting, not the ask. Post this week as a 50-65s talking-head, 100% burned captions, opinionated or number-driven, on a named tool people are already arguing about.
formatstalking-head 5listicle 2tutorial 1
mechanismsrelatable pain 2utility 2novelty 1contrarian 1specific number 1social proof 1
hooksbold claim 6pov relatable 1forbidden or warning 1
contenttalking head advice 4listicle 2reaction 1tutorial 1
topicsproductivity 4tech news 2ai 2ai tools 2startups 2dyson 1artificial intelligence 1gadgets 1
products / appsclaude 3grok 2codex 2dyson camerajet 1chatgpt 1macbook pro 1higgsfield 1claude code 1
Patterns
- Talking-head format with bold-claim hook is the default winning shape 63% of outliers are talking-head at mean 20.5x vs listicle 7.8x; bold_claim hook technique covers 75% of posts at mean 20.1x, e.g. @20vc_tok's Dyson reaction and @sabrina_ramonov's Claude/ChatGPT take.example 1, example 2
- Novelty and contrarian mechanisms massively outperform straight utility Novelty mean 72.4x (@20vc_tok Dyson) and contrarian mean 21.8x (@sabrina_ramonov Claude vs ChatGPT) dwarf utility's 1.1x mean seen in @aistartupfren and @nate.b.jones.example 1, example 2, example 3, example 4
- Naming a specific dollar amount or hard number in the hook lifts performance specific_number mechanism hit 18x mean via @ai.with.andrew's 'websites people are charging $10,000 for' tutorial.example 1
- Naming rival/competing tools directly drives tool-war engagement Claude appears in 38% of outliers (mean 13.6x) and is central to the top two tool-specific hits, both framing Claude against a competitor or a milestone claim.example 1, example 2
- No explicit CTA can outperform asking for comments CTA=none posts average 24.1x vs CTA=comment posts at 9.9x, with the top performer (@20vc_tok Dyson) using no CTA at all.example 1, example 2
Hooks to try
- “The AI tool everyone in SF just quietly switched to this week”
- “Nobody's talking about how [Tool] just made [old workflow] obsolete”
- “I charged $8,000 for a site I built with one prompt”
- “Everyone's dropping [Tool A] for [Tool B]. Here's what they're not telling you”
- “This $499 gadget has more AI than your laptop”
- “The one Claude Code trick that saved my project mid-crash”
- “Stop paying for [Tool] — here's the free version nobody mentions”
- “The startup every VC is fighting to get into right now”
Formats that work
Stick to talking-head, 50-65s, spoken (100% speech share) with burned-in captions (100% share); music is optional (only 38% use it). Tutorials with visual on-screen proof (like @ai.with.andrew's parallax site demo) can outperform standard talking-head when paired with a specific number.
Avoid
- Plain utility/how-to explainers with no strong claim or stakes — mean 1.1x (@aistartupfren, @nate.b.jones)
- forbidden_or_warning hook technique underperformed here at 0.5x (@hackwithantonio 'STOP PAYING FOR AI TOOLS')
- Listicles without a sharp pain point or number stayed mid-pack at 7.8x, weaker than talking-head's 20.5x
- Posting without a named, recognizable tool/brand in the hook — genericized advice pulled the lowest views in the set
Next posts
Tool-rivalry hot take · 60s talking-head, burned captions, no CTA
“Everyone's switching from [Tool A] to [Tool B] and I think that's a mistake”
Mirrors @sabrina_ramonov's contrarian Claude/ChatGPT post (22x) — tool rivalry plus bold claim is proven to spike engagement.
Dollar-amount proof tutorial · 50s tutorial with on-screen build demo, comment CTA
“I used [Tool] to build something I sold for $X — here's the exact prompt”
Matches @ai.with.andrew's specific_number pattern (18x) — concrete money + visual proof outperforms vague advice.
Absurd novelty product reaction · 65s reaction talking-head, no CTA, no music needed
“This [brand]'s new AI [product] does something you won't believe”
Directly replicates the top outlier mechanism (novelty, 72.4x) from @20vc_tok's Dyson toothbrush post — surprising brand+AI combos travel.
Mid-task pain point listicle · 55-60s listicle, burned captions, comment CTA
“[Tool] fails you right when you need it most — here's my 3-step fix”
Follows @ai.with.andrew's relatable_pain listicle (15x) — naming a specific developer frustration plus a structured fix works better than generic tips.
Insider social-proof startup story · 90s talking-head, comment CTA
“This under-the-radar AI startup just became the one every founder is talking about”
Matches @20vc_tok's social_proof post (6.2x) — tech-drama and valuation framing keeps founders' attention even at longer length.
Top posts in the window
| × | handle | hook | format | len |
|---|---|---|---|---|
| 72× | @20vc_tok | Dyson’s new $499 AI-powered toothbrush with a built-in camera | talking-head · novelty | 67s |
| 22× | @sabrina_ramonov | Everyone dropping Claude for ChatGPT 😳 | talking-head · contrarian | 62s |
| 18× | @ai.with.andrew | Claude Fable 5 can now build websites that people are charging $10,000 for. | tutorial · specific number | 49s |
| 15× | @ai.with.andrew | Claude Code hits its token limit right in the middle of your task and you're stuck. | listicle · relatable pain | 58s |
| 6.2× | @20vc_tok | This Silicon Valley startup is the hottest thing in town right now. | talking-head · social proof | 97s |
| 1.2× | @aistartupfren | Use acquire com For startup ideas | talking-head · utility | 29s |
| 1.1× | @evisdrenova | #greenscreen #fyp #xyz | ? | |
| 1.0× | @nate.b.jones | Here's how I use AI to manage my life with agents. | talking-head · utility | 84s |
| 0.5× | @hackwithantonio | STOP PAYING FOR AI TOOLS YOU DON'T EVEN NEED | listicle · relatable pain | 48s |
model claude-sonnet-5 · reports refresh after each scan, or on demand
Outliers of the last 14 days
Posts that reached 5× or more of their own account's baseline (trimmed median views of its last 20 posts), highest multiple first. Engagement outliers ((saves + shares) / views at 3× baseline) are listed after them.
- 200×1.0M vs 4.9k base@ai.with.andrew69k followers14d ago
Claude is offering you a free subscription for one year.
tutorialutilityvideo analyzed · 48seng 1.2× · 38k saves · 13k sharesWhy it worked. Offers immediate financial utility by revealing how to claim $1,000+ in free AI credits with minimal friction.
- 48×267k vs 5.6k base@20vc_tok280k followers4d ago
FRANCE WANTS TO TAX YOUR INTERNET USAGE
talking-headrelatable painvideo analyzed · 75seng 2.0× · 1.4k saves · 2.8k sharesWhy it worked. Tapped into widespread consumer anger over bureaucratic tech taxation and fear of Europe losing tech competitiveness.
- 17×236k vs 5.0k base@bengusberg82k followers11d ago
The rumors are true. Google can now tell you used AI in your writing with 93% accuracy,
talking-headrelatable painvideo analyzed · 76seng 1.0× · 12k saves · 2.3k sharesWhy it worked. Taps into creator and student fear of AI detection, pairing research proof with a quick bypass prompt.
- 16×228k vs 5.0k base@bengusberg82k followers8d agotalking-headutilityvideo analyzed · 60seng 1.2× · 14k saves · 3.2k shares
Why it worked. Leveraged intense job market anxiety paired with specific outcome metrics and a free viral open-source repo.
- 14×60k vs 4.2k base@evisdrenova52k followers10d ago
#greenscreen #ai #anthropic #claude #tech
videoeng 3.2× · 644 saves · 256 shares - 11×68k vs 5.0k base@bengusberg82k followers7d agotutorialutilityvideo analyzed · 54seng 1.4× · 4.8k saves · 1.3k shares
Why it worked. Combines a high-utility Claude hack with simple 3-step execution and instant authority proof.
- 9.7×440k vs 45k base@nate.b.jones498k followers13d ago
If you were wondering why a third of the internet is burning right now, this is why.
talking-headcuriosity gapvideo analyzed · 205seng 1.2× · 5.5k saves · 3.7k sharesWhy it worked. Urgent technical breakdown backed by an exhaustive real-world list of 19 concurrent exploits created intense FOMO and security panic.
- 9.5×54k vs 5.6k base@20vc_tok280k followers9d ago
I would say if you're a pure software investor, you're kind of f**ked, honestly.
talking-headcontrarianvideo analyzed · 37seng 1.4× · 429 saves · 122 sharesWhy it worked. Delivers an aggressive, contrarian thesis attacking the most lucrative venture capital model of the last decade.
- 9.3×52k vs 5.6k base@20vc_tok280k followers2d ago
AI Won't Replace Software Engineers
talking-headcontrarianvideo analyzed · 50seng 1.6× · 497 saves · 125 sharesWhy it worked. Contrarian tech thesis from an established founder directly refutes viral doom narratives about AI developer obsolescence.
- 8.9×8.6k vs 960 base@olegxyz314k followers14d ago
Regular Tools vs AI Alternatives
listicleutilityvideo analyzed · 18seng 1.3× · 200 saves · 74 sharesWhy it worked. Delivers extreme informational density and instant comparison value using a conversational ping-pong split-screen format.
- 7.5×105k vs 5.0k base@bengusberg82k followers7d ago
The rumors are true. ChatGPT is now also watermarking all of its text...
talking-headrelatable painvideo analyzed · 58seng 0.6× · 3.2k saves · 897 sharesWhy it worked. Capitalizes on student and professional fear of AI detection paired with an actionable, technical bypass tool.
- 7.0×39k vs 5.6k base@20vc_tok280k followers7d ago
Ouch. I do not want to be a Google stockholder right now.
talking-headcontrarianvideo analyzed · 84seng 0.5× · 108 saves · 46 sharesWhy it worked. Capitalized on breaking financial panic around a major tech giant with clear contrarian macro analysis.
- 6.7×63k vs 5.6k base@20vc_tok280k followers11d ago
THIS NEW AI VOICE MODEL IS INDISTINGUISHABLE FROM A REAL HUMAN
talking-headnoveltyvideo analyzed · 59seng 2.5× · 1.1k saves · 277 sharesWhy it worked. Anchored breakthrough tech hype to a high-profile tech founder and validated the claim with instant audio proof.
- 5.2×25k vs 4.9k base@ai.with.andrew69k followers6d agotalking-headspecific numbervideo analyzed · 32seng 0.7× · 586 saves · 166 shares
Why it worked. Credibility from a big tech engineer paired with a specific high-percentage cost savings hook drives immediate developer interest.
- 1.8×9.6k vs 5.2k base@sabrina_ramonov865k followers3d agovideoeng 24× · 693 saves · 133 shares
- 1.9×9.7k vs 5.2k base@sabrina_ramonov865k followers6d ago
This AI can blow up your business 🤯
talking-headutilityvideo analyzed · 23seng 15× · 427 saves · 102 sharesWhy it worked. Combines massive social proof stats with a concrete two-tool AI stack and a free downloadable guide incentive.
- 1.3×6.7k vs 5.2k base@sabrina_ramonov865k followers6d ago
Things I'd fix on your resume in under one minute.
talking-headutilityvideo analyzed · 43seng 11× · 232 saves · 30 sharesWhy it worked. Delivers immediate resume fixes paired with a copy-paste AI prompt in under 60 seconds.
- 1.0×5.3k vs 5.2k base@sabrina_ramonov865k followers11d ago
8 YouTube Channels instead of Brainrot 😳
listicleutilityvideo analyzed · 91seng 8.5× · 412 saves · 109 sharesWhy it worked. Taps into guilt around digital doomscrolling by offering immediately actionable, high-status free educational resources.
- 1.2×6.2k vs 5.2k base@sabrina_ramonov865k followers12d ago
Learn Anything So Fast it's Unfair 🤫
listicleutilityvideo analyzed · 58seng 7.7× · 444 saves · 101 sharesWhy it worked. Visual proof of advanced AI outputs paired with actionable, single-phrase commands drives exceptionally high saves and comments.
- 1.4×7.3k vs 5.2k base@sabrina_ramonov865k followers14d agolisticleutilityvideo analyzed · 34seng 7.1× · 482 saves · 116 shares
Why it worked. Demonstrates visually impressive, unconventional ChatGPT outputs that immediately spark curiosity and save-intent for students.
- 0.4×1.8k vs 5.2k base@sabrina_ramonov865k followers13d ago
6 videos to learn the 6 most important life skills.
listicleutilityvideo analyzed · 33seng 6.2× · 102 saves · 30 sharesWhy it worked. Curates high-value, familiar educational YouTube topics into a single quick-reference listicle driving high save rates.
- 1.8×9.5k vs 5.2k base@sabrina_ramonov865k followers13d ago
This is the one prompt you need to make AI sound less robotic and more like you.
talking-headutilityvideo analyzed · 27seng 6.0× · 620 saves · 42 sharesWhy it worked. Solves a universal pain point in generative AI writing by offering a complete swipeable ruleset with a comment-gated lead magnet.
- 4.2×22k vs 5.2k base@sabrina_ramonov865k followers12d ago
This is the one thing you should do before every interview.
tutorialutilityvideo analyzed · 25seng 5.2× · 1.1k saves · 255 sharesWhy it worked. Delivers an immediately actionable, copy-pasteable AI prompt addressing universal job interview anxiety.
- 3.4×153k vs 45k base@nate.b.jones498k followers11d ago
AI Hack: The Prompt That Scans You For Vulnerabilities
talking-headutilityvideo analyzed · 93seng 5.2× · 10k saves · 3.5k sharesWhy it worked. Provides an immediate, high-utility defensive AI prompt resolving heightened viewer anxiety around hacking.
- 1.7×8.5k vs 5.2k base@sabrina_ramonov865k followers4d ago
if you have zero friends like me...
videoeng 5.2× · 141 saves · 20 shares - 1.6×8.1k vs 5.2k base@sabrina_ramonov865k followers4d ago
What salary question trips you up?
videoeng 5.0× · 122 saves · 24 shares - 0.5×2.3k vs 5.2k base@sabrina_ramonov865k followers7d agotalking-headcontrarianvideo analyzed · 19seng 4.6× · 98 saves · 26 shares
Why it worked. Targeted widespread job seeker frustration with ghost listings by introducing a clean AI-filtered alternative.
- 1.1×4.5k vs 4.2k base@evisdrenova52k followers7d agotalking-headcontrarianvideo analyzed · 96seng 4.4× · 70 saves · 22 shares
Why it worked. Delivers an authentic, high-conviction motivational monologue addressing widespread tech anxiety with personal cross-disciplinary credibility.
- 0.3×1.4k vs 5.2k base@sabrina_ramonov865k followers7d agovideoeng 3.8× · 51 saves · 11 shares
- 3.0×2.9k vs 960 base@olegxyz314k followers12d ago
These five AI certificates will get you hired in 2027.
listicleutilityvideo analyzed · 38seng 3.7× · 209 saves · 50 sharesWhy it worked. Capitalizes on career anxiety and AI upskilling urgency with high-value curation.
Accounts in this niche
15 TikTok accounts, largest first. Baseline = trimmed median views of the account's last 20 posts. Is one of them yours? Ask us to leave it out.
- @sabrina_ramonov865k followers · 5.2k baseline
- @samdespo721k followers · 9.1k baseline
- @nate.b.jones498k followers · 45k baseline
- @olegxyz314k followers · 960 baseline
- @20vc_tok280k followers · 5.6k baseline
- @justyn.ai182k followers · 5.5k baseline
- @brycent_156k followers · 11k baseline
- @hackwithantonio126k followers · 1.3k baseline
- @webloved108k followers · 7.7k baseline
- @swerikcodes93k followers · 15k baseline
- @bengusberg82k followers · 5.0k baseline
- @khris.sheer70k followers · 805 baseline
- @ai.with.andrew69k followers · 4.9k baseline
- @evisdrenova52k followers · 4.2k baseline
- @aistartupfren32k followers · 1.5k baseline
About this report
What counts as an outlier here?
A post from the last 14 days that reached at least 5× its own account's baseline. The baseline is the trimmed median views of that account's last 20 posts, so a breakout is measured against the account itself, not against the niche.
How often is this report refreshed?
After every scan of the niche; the last one ran 8h ago. Posts younger than 48 hours show as climbing and are never flagged, so the list settles rather than churns.
I'm one of these creators. Can I be left out?
Yes. Email support@ninefold.ai with your TikTok handle, or use the removal link under the accounts. We stop tracking the account and delete the posts, counts and analyses we hold about it within a month, and it comes off every public page. These pages show short hooks, public counts, links and our own analysis; no creator photos or videos are stored or served.
How do I use this niche?
Copy it into your account for free, then scan it on your plan to get the current outliers, a teardown of each (hook, format, structure, why it worked) and three adaptations in your own voice. From Claude or ChatGPT, call use_public_niche with the slug.
From Claude or ChatGPT with the Outliers connector: use_public_niche with ai-founders-15ef81. See the docs.