Nate Herk YouTube Breakdown: 5 Lessons for AI Tool Creators

Nate Herk's Claude Code course reached about 1.2 million views, or 9.2 times the channel's nearby normal. That made me curious about how he packages a large learning promise. I measured public view counts of every upload in the last 2 years, as of November 8, 2026, then read first-minute transcripts and sampled frames. Here are the teaching choices I'd borrow for my own AI and coding videos.

What stood out

The top-five ranking covers qualifying videos from the last 12 months, ordered by performance against each upload's local normal. The wider research window supplies the lesson comparisons.

Video	Views, approximately	Times normal
Claude Code course	1.2 million	9.2
n8n start-over guide	344,000	6.3
Beginner Claude Code guide	461,000	5.8
Karpathy Claude Code video	734,000	5.7
Website tutorial	588,000	5.1

Claude Code keeps returning, and the teaching titles describe a destination. My takeaway is to show the result clearly, then teach a path someone can follow.

1. Show your own build

Titles saying what Nate did reached about a 38% hit rate, against about 15% for other titles. A hit means at least double local normal. This comparison contains 60 titles against 291 others; topics and thumbnails overlap.

In the August 22nd website design skill video, he opens with a finished website beside his facecam, then shows the original and redesign. That individual example is an anecdote. I like the teaching move because the viewer can inspect the result behind the personal claim.

I'd try: I built or tested, then the specific thing, then what you'll learn. Record the working output and the attempt that failed. A cooking version could show cut-open roast potatoes after testing preparation methods.

2. Keep the useful depth

Uploads at least 20 minutes long reached about a 31% hit rate, compared with about 13% for the shorter group. That's 129 longer uploads against 222 shorter ones, with different topics and course promotion. Extra minutes alone aren't a recipe.

The January 21st beginner Claude Code guide opens with an agenda on a white canvas: interface, planning, testing, and publishing the result. The route is visible before the viewer starts following it.

I'd outline setup, build, troubleshooting, and the finished result before recording. Keep the installation, the error message, and the check that proves the project works. Cut repetition around those steps. A gardening tutorial could retain the drainage check in a balcony salad planter build so viewers can test their own setup.

3. Share the working resource

Titles explicitly offering a free template reached a 40% hit rate, against about 18% for the other group. All 15 template titles are from 2025, compared with 336 others. Timing overlaps, and downloads weren't measured.

The design skill video offers a visible example of connecting a resource to a result. Nate offers the reusable instructions free while the finished website is on screen. Its title calls the resource a skill; the free offer appears in the opening. Keep that distinction in mind when connecting this example to the title comparison.

I'd share the resource used on camera, test it from a clean setup, and include instructions and required accounts. For woodworking, that could mean the demonstrated shelf's cutting plan, with material thickness and a print-scale check.

4. Name the beginner barrier

No-code titles reached about a 36% hit rate, against about 18% for other titles. The groups contain 28 titles and 323 others. Every no-code title predates 2026, so timing overlaps; ease wasn't measured.

The February 19th website tutorial shows a useful beginner sequence. Nate reassures new users, shows where to download the editor and install the Claude Code extension, and explains the account requirement. The promise has a visible starting point.

I'd name the desired result and the specific barrier removed, then show the accessible route and its remaining requirements. For a home strength routine, that means showing an easier bodyweight option and the space needed to follow along.

5. Build a topic lane

Within 2026, titles naming Claude Code reached about a 33% hit rate, against about 16% for other titles: 61 against 142. Tool demand and release timing overlap, and the literal title filter misses relevant videos without the name.

In the April 5th Karpathy video, Nate starts with an Obsidian graph, then opens a note containing a source link, summary, and takeaways. Claude Code has a concrete job: organizing videos into a searchable knowledge system. Website building gives the same tool another job.

I'd plan related tutorials for the same audience and tool, each with a distinct outcome. Compare results after the age cutoff before extending the sequence.

How I measured

I used public upload view counts as of November 8, 2026. The research includes 355 videos outside Shorts. Removing uploads under 14 days leaves 351 for lesson comparisons. Shorts stay separate.

Local normal is the middle view count of the 20 nearest other qualifying uploads by date, before and after a video. Divide the video's public view count by that baseline to get times normal. For the March course, normal was about 133,000 views.

The opening transcripts and sampled frames support observations about teaching choices. These comparisons mix topics, timing, packaging, and audience demand. My recommendations are opinions to test on your own channel.

My checklist

1. Test a real build title.
2. Keep setup and troubleshooting.
3. Share the working resource.
4. Name the beginner barrier.
5. Keep the tool recognizable while changing the job.

Which lesson would you test on your next upload, and what would that video be about?

Bring your own channel questions to the free 500x Engineer community.