Nate's unfair advantage is being first, in public, on camera
Most creators wait a week to form an opinion on a new AI tool. Nate opens it, tests it, and posts the result before the news cycle moves on.
Nate Herkelman (known to his community as Nate Herk) is the founder and CEO of Uppit AI, the education platform behind AI Automation Society, a community that has grown past 400,000 members. His LinkedIn feed is not a highlight reel of hot takes. It is a near-daily log of him opening whatever Anthropic, OpenAI, or the open-source scene shipped that week, running it on a real task, and posting the result while it is still news.
That is the whole engine. Same-day-teardown growth is when a creator turns their own first real test of a new AI tool into a post, published within a day of the launch, before opinion has settled and while the search demand is still climbing. Nate runs it on a punishing schedule: install the tool, ship something real with it, name the exact cost and the exact caveat, then point the reader to a longer YouTube build for the depth.
A rundown of specs copied from the changelog, forgotten by the next scroll.
The same launch, tested live on a real task, with a real cost and a real caveat attached.
“The model matters less than the system you wrap around it.”
— From his most-reacted post, on shipping a full YouTube video with Claude Fable 5 while he was at the gym (1,933 reactions)
Five findings that repeated across 100 posts
- His biggest posts are same-day proof, not opinion: the Claude Fable video post (1,933 reactions), a video-editing breakdown (1,457), and a post built on Karpathy's Claude Code wiki idea (1,402).
- Volume is the edge. He posts about 7.2 times a week, 113 own posts in 111 days, the heaviest cadence in this playbook series.
- Comments run hot for a technical account: a 12.7% comment-to-reaction ratio, roughly double the ~6% LinkedIn norm.
- Video carries the account. 69% of his posts are video (78 of 113) and they hold every spot at the top of the leaderboard.
- Real reach still happened: 6 of 113 posts cleared 1,000 reactions, on an account built around process over virality bait.
The numbers behind the account
About 7 posts a week, heaviest on Tuesday and Friday, with video doing almost all of the reach.
Across the 100 posts we analyzed, Nate published roughly 7.2 times a week, well above the 4-post-a-week rhythm we see from most founder accounts, which lines up with how the platform rewards a steady daily signal, something we break down in our guide to how the LinkedIn algorithm works.
When he posts
The content-type mix
Where the engagement comes from
The top posts
| # | Post | Reactions | Comments | Reposts |
|---|---|---|---|---|
| 1 | Claude Fable 5 Made This Entire Video By Itself | 1,933 | 243 | 120 |
| 2 | Claude Just Changed Video Editing Forever | 1,457 | 86 | 105 |
| 3 | Andrej Karpathy Just 10x'd Everyone's Claude Code | 1,402 | 136 | 86 |
| 4 | Today I was recognized in a Forbes article | 1,356 | 211 | 13 |
| 5 | Fable 5 + Karpathy's LLM Wiki is Basically Cheating | 1,174 | 114 | 66 |
| 6 | Claude Video Editing Just Became Unrecognizable | 1,055 | 139 | 58 |
Want to see how your own account stacks up on cadence and engagement? Run it through our free LinkedIn analyzer.
The six content pillars
Every post is one of six repeatable buckets, which is how a single creator keeps up with a news cycle that moves every day.
A raw script or prompt turned into a finished, edited video with Claude, HeyGen, and ElevenLabs.
His recurring 4 Cs framework, Context, Connections, Capabilities, Cadence, for running a business out of Claude Code.
Milestones for AI Automation Society, press mentions, and results members post about him.
Same-day reactions to a new model or feature, tested on a real task first.
Claude Code vs Codex, Opus vs GPT, GLM vs Opus, on identical prompts.
Frameworks for selling AI work and turning automation skills into income.
Pillar 1: Same-day AI video production (the reach engine)
Why it works: His single biggest post is a receipt, not a claim. He names the exact tools and the exact absence of himself from the process. A finished, verifiable result beats a promise.
Pillar 2: The AI Operating System, second brain (the framework)
Why it works: He borrows the idea from a named authority, then shows his own working version. Naming the source builds trust; showing his own build proves he actually used it.
Pillar 3: Community and authority proof (the credibility)
Why it works: Third-party proof outperforms almost everything he could say about himself, and naming the people behind the milestone turns a humblebrag into a thank-you.
Pillar 4: Breaking Anthropic and AI product news (the volume)
Why it works: He never just reports the launch, he lists exactly what he personally tested that same day. That numbered test list is the steady drumbeat behind an almost-daily cadence.
Pillar 5: Head-to-head model and tool tests (the proof)
Why it works: He gives both tools a real win and a real loss instead of crowning one. Honest, specific comparisons earn comments that vague hype posts never do.
Pillar 6: Career and business advice for AI builders (the ladder)
Why it works: The pyramid is a fill-in-the-blank framework any AI builder can reuse to explain their own pricing, and it earned a 22% comment ratio, the highest of any pillar.
The hooks that earned the click
The through-line is proof over promise. Nate's first line is almost always a result he already has in hand.
State the shift as fact. "Claude Just Changed Video Editing Forever."
Open on walking away and coming back to a finished result. "...went to the gym, and came back to a finished video."
Borrow credibility from someone the audience trusts. "Andrej Karpathy Just 10x'd Everyone's Claude Code."
Open with a real dollar figure. "I Just Spent Over $400 on Claude Design Usage."
Promise a specific count of wins. "32 Claude Code Hacks (Beginner to Pro)."
Correct a common belief in the first line. "Most clients aren't actually buying AI systems."
For the mechanics of writing openers like these, our guide to writing LinkedIn hooks goes deeper, and you can pressure-test your own first line in the free hook generator.
His top hooks, by the numbers
| Hook type | Opening line | Reactions |
|---|---|---|
| Hands-off build | "I opened Claude, gave it one goal, went to the gym, and came back to a finished video." | 1,933 |
| Named-authority drop | "Andrej Karpathy Just 10x'd Everyone's Claude Code." | 1,402 |
| Stat-lead | "I uploaded 29 YouTube videos in the past 28 days." | 876 |
| Confession-spend | "I Just Spent Over $400 on Claude Design Usage." | 601 |
A voice built for a technical, skeptical audience
It reads like a builder showing his work, numbered steps, real costs, and an honest caveat before anyone can call it hype.
- States the shift as fact in the title, then earns it in the body.
- Numbers every step. Most posts run as a 1️⃣ 2️⃣ 3️⃣ list, never a wall of prose.
- Names the real stack and the real cost, down to the dollar and the model version.
- Volunteers the honest caveat, even inside his best-performing posts, not just his worst.
- Pushes depth to the comments. Almost every post ends with "Link to the full video is in the comments."
The voice is recognizable partly because of recurring devices: the numbered-emoji list (1️⃣ 2️⃣ 3️⃣) for any process, an arrow (→) for a sub-point, and a closing line naming the length of the companion YouTube video.
What he does, and doesn't, do
- Post the tool test the same day it drops
- Open on the finished result, not the setup
- Name the real stack down to the version number
- State a real cost or a real caveat, even an unflattering one
- Push the tutorial depth to a linked YouTube video
- Burying the news two paragraphs in
- Selling the feature list instead of a tested outcome
- Hiding the real cost or the failure case
- Posting a teaser with no way to go deeper
- Writing in one dense, unbroken paragraph
Holding that voice across a news cycle that reshapes itself weekly, at 7-plus posts a week, is the part almost nobody sustains, and it is exactly the gap CaptureFlow closes. CaptureFlow is an AI content agent that turns your expertise into weeks of on-brand content for every platform. You capture one idea in 5 minutes (a voice note, a screen recording of the tool you just tested, a raw take), and CaptureFlow, trained on your voice and your past posts, drafts native content for each channel, a LinkedIn post, a carousel, a short video, so a daily cadence never depends on finding an hour to write. See how the AI content agent works.
The systems underneath the posts
A funnel and a flywheel quietly turn 100 posts into YouTube views, community members, and press.
The LinkedIn-to-community funnel
LinkedIn is the top of the funnel, not the destination. Every post routes to a longer video, then a free resource, then the community itself.
The proof flywheel
- 1Nate tests a new tool the same day it shipsA real task, a real cost, a real caveat.
- 2The post earns reach and unusually high comments12.7% comment-to-reaction ratio.
- 3Community members ship their own resultsTestimonials get reposted and quoted.
- 4Outside proof compounds the authorityA Forbes mention, a Google Knowledge Panel, 400,000 members.
- 5The authority earns the benefit of the doubt on the next testThe cycle starts over.
Choosing the media
A screen recording of the actual build, cold-opened on the result.
A real screenshot of the article, the panel, or the member-count graph.
Plain text. The pyramid or the numbered list carries the post without a visual.
This news-cycle model sits close to what we mapped in the Ruben Hassid playbook, another AI creator whose growth runs on giving away the real process, and it is the template worth studying for anyone building an audience of creators in a fast-moving niche.
Your 30-day challenge
Run the playbook for a month. Turn the tools you're already testing into proof, one pillar at a time.
- Days 1-2: Pick the newest tool in your niche and test it on one real task
- Days 3-4: Post the result the same day, with the exact stack and cost named
- Days 5-7: Record a longer walkthrough and link it in the comments
- Days 8-9: Turn your recurring process into a named framework (your own '4 Cs')
- Days 10-11: Post one honest comparison, tool A versus tool B, on identical prompts
- Days 12-14: Share one real cost or failure case from the past month
- Days 15-17: Highlight a member or customer result, named, with the number attached
- Days 18-19: Post a milestone (users, members, revenue) with the honest context behind it
- Days 20-21: Write a career or business-advice post that reuses your own pyramid or ladder
- Days 22-24: Re-test last month's tool against this month's new release
- Days 25-27: Repost a testimonial or proof point from your own community
- Days 28-30: Review which pillar earned the highest comment ratio and double down
Want the daily cadence without writing every teardown from scratch? That is exactly what CaptureFlow's content agent automates, and you can see current plans on our pricing page.
The metrics to track weekly
| Metric | Benchmark to aim for |
|---|---|
| Reactions per post | 300+ |
| Comments per post | 30+ |
| Weekday posting cadence | 5+ per week |
| Comment-to-reaction ratio | 10%+ |
| Reposts per post | 20+ |
| Named community wins per month | 2+ |
The takeaways
- 01Post the same day the tool drops. Nate's biggest posts are teardowns published within a day of a real launch, before the news cycle moves on.
- 02Lead with the finished result, not the setup. His top post opens with him walking away and coming back to a done video.
- 03Let the comments carry the depth. A 12.7% comment-to-reaction ratio, roughly double the LinkedIn norm, comes from routing readers to "the full video is in the comments."
- 04Name the real cost and the real caveat. Dollar figures and honest limitations earn trust that a pure feature list never does.
- 05Post almost every day. 7.2 posts a week, the highest cadence in this playbook series, is what compounds a fast-moving niche.
- 06Batch-capture your reactions so a near-daily cadence doesn't mean writing every post from scratch.
Frequently asked questions
- How did Nate Herkelman grow his LinkedIn following?
- By testing new AI tools the same day they launch and posting the real result, cost, and caveat, almost every day. Across 100 recent posts he averaged 403 reactions each, and his AI Automation Society community has grown past 400,000 members.
- What kind of post performs best for Nate Herkelman?
- A same-day teardown of a new AI tool with a finished, verifiable result. His top post, an entire video made by Claude Fable 5 while he was at the gym, earned 1,933 reactions, and a post built on Andrej Karpathy's Claude Code wiki idea earned 1,402.
- How often does Nate Herkelman post, and when?
- About 7.2 times a week, the heaviest cadence in this playbook series, with Tuesday and Friday leading but posts landing almost every day including weekends.
- How do you apply this playbook without spending hours a week?
- Batch-capture your reaction to whatever you're testing that week, then let a content agent draft in your voice. CaptureFlow turns one 5-minute capture into a week of native posts across platforms, so you can hold a near-daily cadence without writing every post from scratch.
