AI Side Hustles Half-true — works only if you do the unspoken work

By the Income Reality Check Editorial Team · Updated 2026-10-17 · 7 min read

Nate Herk’s 10 Claude GitHub repos: the tools are free, but who’s paying you?

Verdict: Half-true — works only if you do the unspoken work. The repos are real and genuinely good; the income lives entirely in the client work the video treats as an afterthought.

Nate Herk’s video “I Tried 100+ Github Repos for Claude. These 6 Are the Best” has pulled in more than 238,000 views on his channel, Nate Herk | AI Automation. The pitch: install a handful of free, open-source tools for Claude and you can “build something that businesses will actually pay you for” — videos, websites, apps, SEO audits — with no prior experience in any of it. The tools are real. The part that turns them into money is not nearly as simple as he makes it sound.

What the video actually claims

The headline isn’t a dollar figure. It’s a promise about leverage. Herk walks through ten GitHub repos — Hyperframes for video rendering, a Google Workspace CLI, the Apify scraper connector, his own Scrollcraft design engine, Humanizer for copy, Addy Osmani’s Agent Skills, Expo’s mobile-app skills, Supabase and Stripe best-practice skills, Playwright for testing, and Open SEO — and frames each one as a service you can sell.

A gym wants 20 captioned clips a month? Sell a content package. A cleaning company wants online booking and deposits? Build the app. A brand wants a custom scroll site? Designers “still pay real money” for that. His closing instruction is the whole business model in two sentences: “Pull a list of businesses with no website, build one off a sample page, clean up the copy, and then send the link.” He’s blunt that the repos themselves aren’t the product — “No one’s going to pay you for a GitHub repo, those are free” — and that you’re selling what you build around them.

To his credit, the video is unusually honest about the tools. He flags open bugs, warns that some token-saving repos (Caveman, RTK, Ponytail) massively underdeliver versus their advertised savings per JetBrains testing, and tells beginners not to install everything at once. This isn’t a scam pitch. It’s a competent tool review with a service-business thesis stapled to the end — and the thesis is where the gaps are.

What the method actually requires

Here’s the move the video makes and almost every viewer misses: it spends 20 minutes on production (the repos) and about 15 seconds on distribution (finding and closing clients). That ratio is backwards relative to where the difficulty actually sits.

Consider the “send the link” cold-outreach plan. Finding businesses, qualifying them, writing a message that doesn’t get deleted, following up, handling objections, pricing, scoping, contracts, revisions, invoicing, chasing late payment — none of that is in the repos, and all of it is the job. The U.S. Small Business Administration’s own guidance on finding customers describes market research and competitive analysis as deliberate, often time-consuming work: surveys, interviews, demographic data, figuring out where your buyers already go (sba.gov). That’s the foundation under a single cold email — not a step you skip.

Then there are the costs the video mentions in passing and never totals up. Some are small; some recur.

None of these are dealbreakers. The point is that “completely free” applies to the repos, not to running a real client project. An app build alone starts at $99/year before you’ve earned a cent.

And the skill floor is higher than “no experience needed” implies. Herk says you don’t need video editing experience to use Hyperframes — fine for making a clip. But delivering reliable work a business will pay for repeatedly, debugging Supabase row-level security, verifying a Stripe webhook actually came from Stripe, shipping an app through App Store review? That’s closer to the work the Bureau of Labor Statistics tracks under web developers, where the median wage was $92,650 in May 2025 and the bottom 10% still earned about $48,100 (bls.gov). People get paid that because the delivery is hard, not because the tooling is scarce.

Who actually wins this game

Look at who Herk is. He runs an AI automation channel with a large audience, an “AI Automation Society,” event sites, and a team he references casually (“me and my team”). He has distribution. When he says “send the link,” he’s describing the last 5% of a funnel he already built — the 95% being the audience and reputation that make a cold pitch land.

The people who turn these repos into income fall into familiar buckets: operators who already have a client book or an audience to market to, freelancers with existing domain skills who are now faster, and early movers in a specific local niche before it fills up. A beginner with the same ten repos and zero pipeline has the tools and none of the demand. Would a bakery hand its booking system to a stranger who emailed a link last Tuesday? Occasionally. Reliably enough to pay rent? That’s the question the video doesn’t ask.

What you’d realistically earn

There’s no promised number to debunk, so measure against the implied one — a sellable service business. A realistic beginner trajectory looks like months of unpaid pipeline-building: outreach, free samples, a first small gig in the $200–$1,000 range if you’re persistent and your niche is well-chosen. Steady income — a few hundred to a couple thousand a month — tends to arrive after you’ve landed repeat clients and referrals, which is quarters of consistent work, not an afternoon.

That’s not a knock on the tools. It’s the normal shape of a service business. Freelance web and automation work can genuinely pay well; BLS pegs the field’s median near six figures. But that money reflects accumulated clients, reputation, and delivery skill — exactly the assets a brand-new operator hasn’t built yet, and exactly the ones a GitHub repo can’t install.

Who this is (and isn’t) for

This path fits you if you already have some client-facing muscle — you’ve freelanced, sold something, or have an audience — and you want to ship faster and move upmarket into apps and SEO retainers. Budget $100–$300 to start (Apple account, a scraper plan, SEO trials) and expect to spend most of your hours on sales and delivery, not on Claude. It’s a poor fit if you’re hoping the repos replace the business itself, have no time for outreach, or expect passive income. There’s nothing passive here. The tools compress production; they don’t generate demand.

What to remember

Nate Herk’s repo list is one of the more honest tool roundups you’ll find, and the ten picks are legitimately useful. The video just quietly swaps “here are great tools” for “here’s a business,” and those aren’t the same claim. The tools are free and real. The income depends on the sales, scoping, and delivery work the video compresses into one closing line — and that work is the actual job. (Worth knowing: the FTC has proposed expanding its Business Opportunity Rule to cover money-making and coaching programs that make deceptive earnings claims, so U.S. readers should treat any “no experience, guaranteed income” framing with care — see ftc.gov.)

For more on where Claude-based services actually produce income, see our looks at Claude AI side hustles that can pay a full-time income and how AI can run a content business.