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

By the Income Reality Check Editorial Team · Updated 2026-05-31 · 7 min read

Claude Code as an ‘AI operating system’: the costs the video skips

Verdict: Half-true — works only if you do the unspoken work. The “context is king” insight is genuinely good; the part where it quietly saves you money and runs your business is where the pitch thins out.

Nate Herk’s video “I Turned Claude Opus 4.8 Into My Entire AI Operating System” (Nate Herk | AI Automation, ~83,000 views) makes a confident claim: stop bouncing between Chrome tabs and SaaS tools, route everything through Claude Code, and you get a “second brain” that runs all your businesses. He even calls it cheaper, because you cancel the other subscriptions. Is the idea real? Mostly — but the version that works costs more, takes more, and risks more than the calm narration lets on.

What the video actually claims

Herk’s framing is a “default shift.” Because Claude Code runs on the same underlying model as the Claude chat app (Opus 4.8, in his telling), he argues you should reach for the terminal first — even for brainstorming, writing, or planning that has nothing to do with code. The payoff he describes is consolidation: one place that reads his meeting transcripts, Slack threads, ClickUp tasks, email, YouTube transcripts, and QuickBooks figures, so it can answer “what should I do next week?” with real context instead of generic filler.

He builds the pitch on two frameworks — the three M’s (mindset, method, machine) and the four C’s (context, connections, capabilities, cadence) — and offers a free GitHub repo plus a free three-hour course inside his Skool community to get you started. The thesis line is the memorable one: “Context is king, not the AI model.” Everyone has the same Opus 4.8, so your edge is the fuel you feed it.

Notice what the video does not promise. There’s no “$10,000 a month” number, no “passive income,” no income method you copy and resell. That honesty is worth crediting. But it also means the value proposition for a viewer hangs on three quieter claims: that this is cheaper than your current stack, that you can replicate his setup, and that the autonomy is safe enough to trust. Each deserves a closer look.

What does running everything through Claude Code actually cost?

Start with the money, because “potentially cheaper” is the slipperiest line in the video. Claude Code is token-hungry. Anthropic’s own help center notes the lowest paid tier (Pro, $20/month) includes Claude Code, but heavy users “hit their usage limits much faster” than they would in the web chat. Routing your entire working day — brainstorming, writing, planning, plus actual coding — through it is exactly the heavy-use pattern that blows past Pro.

So what does the real version cost? Most people living in Claude Code the way Herk describes end up on the Max plan. Per Anthropic, Max runs $100/month (5×) or $200/month (20×), with weekly usage limits that reset seven days after a session starts. If you’d rather pay per token through the API, Opus 4.8 is priced at $5 per million input tokens and $25 per million output tokens — and agentic sessions that re-read large context files burn input tokens fast. The “I cancelled my other tools” savings are real, but they’re often replaced by a $100–$200 monthly bill, not zero.

Then there’s time, the cost nobody invoices you for. Herk says some skills take “50 tries” before they’re good, that he edits his configuration files “almost every day,” and that he reorganizes projects weekly. He’s candid that adopting a new automation creates a “20% dip” — a stretch where the AI workflow is slower than doing it by hand — before it pays off. That dip is real labor. Building, testing, and maintaining a personal AI operating system is a part-time job layered on top of your actual job.

And the third cost is risk. Herk tells a striking story: an agent on his own team “proactively picked up a to-do list” and sent three promotional emails to over 150,000 inboxes that were never meant to go out. They took down a page and apologized. He’s right that the fix is scoping permissions — don’t put a “send email” key on the ring if you don’t want it used — but the lesson generalizes. This isn’t hypothetical industry-wide, either. Anthropic itself warns that Claude’s computer use “is still early” and “can make mistakes”, and CNBC has reported the company’s own CEO flagging a security “moment of danger” as autonomous AI exposes vulnerabilities at scale. More reach means more ways to be wrong, fast.

Who actually wins this game

Here’s the part the framing skips: an AI operating system organizes a business. It doesn’t create one.

Herk’s setup is valuable precisely because he already has meeting transcripts, a team, QuickBooks revenue, a YouTube channel, and a Skool community for it to read. Feed Claude Code all of that and “what should I do next week?” produces something useful. Feed it an empty folder and it has nothing to be king of.

So the people who win with this are operators who already have data exhaust and recurring workflows — a freelancer juggling ten clients, a small agency, a creator running several income streams. For them, consolidation is a genuine multiplier. The people who don’t win are beginners hoping the OS itself is the income. There’s no money method here to copy; the video is a productivity philosophy aimed at someone who’s already busy. The free repo and free course, meanwhile, funnel toward what Herk calls the largest AI automation community on Skool — a classic lead magnet. Nothing shady about that, but it’s worth seeing the funnel for what it is.

What you’d realistically earn

This is where an honest reality check has to push back on the category, not just the video. Because no income figure is promised, there’s no “$10k/month” to debunk — and that’s a tell. A tool that organizes your work has no payout of its own.

If you already run a profitable operation, a well-built AI OS might save you several hours a week and tighten your decisions; converted to dollars, that’s real but indirect and entirely dependent on your existing business. If you don’t have a business yet, your realistic “earnings” from building an AI operating system in the first six months are roughly the same as your earnings from buying a very nice label maker: $0, plus a tidier desk. The U.S. FTC’s pending rule changes on deceptive earnings claims would require sellers of “business coaching or investment opportunities” to substantiate income claims on request — a useful reminder (for U.S. viewers especially) to be skeptical whenever a free course implies a financial outcome it never quite states. Herk mostly avoids that trap; many videos in this genre don’t.

Who this is (and isn’t) for

This makes sense if you already run a business or several income streams, you have 3–5 hours a week to build and maintain skills, you’re comfortable in a terminal, and you can absorb a $100–$200 monthly subscription plus the occasional agent mistake. If that’s you, the four C’s are a sane way to think about it, and the consolidation is real. It does not make sense if you’re looking for a side hustle that pays — there isn’t one here — or if you’d be wiring a credential-loaded agent into live email and money systems before you’ve learned to scope its permissions. Start narrow, watch it like you’d watch a kid on a bike, and earn the autonomy.

What to remember

Herk’s core insight holds up: your context, not the model, is the edge, and pulling your scattered work into one place is a legitimate productivity gain. The half-truth is “cheaper and it runs your business.” It’s usually a $100–$200/month tool plus ongoing build time plus real risk — and it amplifies a business you already have rather than handing you one. Treat it as an upgrade to how you work, not a route to income, and the pitch lands honestly.

For more on where Claude actually does and doesn’t generate revenue, see our reviews of the “$20,000/month with Claude Code on YouTube” claim and starting a one-person business with Claude AI in 30 days.