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

By the Income Reality Check Editorial Team · Updated 2026-08-06 · 7 min read

Manus AI built a whole app in one session — but not a business

Verdict: Half-true — works only if you do the unspoken work. The demo is real; the “app launch infrastructure” it claims to produce is a working prototype, not customers or income.

Shark Numbers’ host stops “testing features” and hands Manus AI a single job: take an app idea from a blank prompt to something shippable. In one session — narrated as roughly an hour — Manus researches the market, builds a mobile app called Scroll Cost, drafts a list of 50 creators to pitch, and then drives a live browser to shop Amazon. The pitch underneath is seductive: the gap between having an idea and having “something real to show” has collapsed “from weeks to hours.” Is that true? Mostly, at the production step. The part the video waves past is everything that turns a built app into money.

What the video actually claims

The host picks the highest research tier (“Manus 1.6 max”), runs a “wide research” job across the App Store, Google Play, Product Hunt, and five Reddit communities, and gets back a structured market-gap report. The conclusion: nobody has built an app that converts your daily screen time into money lost based on your hourly rate. People calculate it by hand on Reddit; no dedicated app exists. “The gap is real,” he says.

From there, Manus builds the app from its own research PDF — five build tasks, a live preview, and an auto-generated logo made with GPT image 2 that “wasn’t part of the prompt.” He enters a $50 hourly rate, adds two hours each of Instagram and TikTok, and the app reports $200 lost today, $73,000 a year. Then Manus produces a 50-creator outreach list (17 Instagram, 17 YouTube, 16 TikTok) with handles, follower counts, and contact emails, and offers to draft personalized emails and connect Gmail and Meta Ads Manager. It finishes by operating a real browser to find a phone-blocker on Amazon.

The framing is the claim: “a complete app launch infrastructure built in one session.” No dollar income is promised outright. The implied payoff is that you, too, can compress an entire product launch into an afternoon.

What does one session actually cost?

Start with the meter, because the video mostly hides it. Manus runs on credits, and the tiers in 2026 are $0 (free, 300 daily credits), $20/month (4,000 credits), $40/month (8,000), and $200/month (40,000), per pricing breakdowns from Lindy and others. That works out to roughly half a cent per credit. Sounds cheap. It isn’t, once you see consumption.

A single deep-research task can burn 900 to 1,000+ credits, and the loudest complaint across Reddit and Trustpilot is that Manus often doesn’t reliably tell you the cost before you commit — and that failed tasks still eat credits. The video shows three heavy jobs (wide research, an app build, a second wide research) plus browser operation. On the $20 plan’s 4,000 monthly credits, a workflow like the one on screen can swallow a large slice of the month in one sitting. (The host claims Manus showed the cost up front; many users report the estimate is inconsistent. Budget for the pessimistic case.)

Then there’s shipping. “Publish directly to Google Play” and “connect a developer account” for iOS skip the accounts you have to buy and maintain:

None of that is a dealbreaker. But “one prompt, a working mobile app with publishing options” quietly becomes a few hundred dollars a year before a single user downloads anything — and app-store review, privacy disclosures, and store listings are still your problem, not Manus’s.

Who actually wins this game

Here’s the sleight of hand common to AI-agent demos: they compress the production step, which was never the bottleneck. Anyone can build an app. Getting people to install it and pay is the hard part, and Manus doesn’t do that for you.

Look at the outreach list. Fifty creators with contact emails is genuinely useful — that’s real leg-work automated. But a list is not a campaign. Cold-pitching productivity influencers has a low single-digit reply rate at best, and the ones with real audiences get pitched constantly. Manus offering to “run paid promotion” through Meta Ads Manager is the honest tell: the growth still costs money, and it’s your money. The people who win with a tool like this already have a distribution edge — an existing audience, a paid-traffic budget, or a niche where they know exactly who to reach. For someone starting cold, Manus shortens week one and does nothing for weeks two through fifty-two. We made the same point about content tools in AI can now run your entire content business — here’s how.

What you’d realistically earn

The video shows $200/day and $73,000/year — but those are numbers inside the demo app, describing a hypothetical user’s screen-time “loss.” They are not earnings from the app. It’s worth being blunt about that, because a skimming viewer could easily conflate the two.

Real revenue from a brand-new consumer app is close to zero for months. Most indie apps never clear meaningful money; a majority earn under a few hundred dollars total in their first year, and the median is effectively nothing without marketing spend. A screen-time novelty app faces a crowded category (Opal, one-Sec, ScreenZen, plus Apple’s built-in Screen Time) and no obvious reason for someone to pay. A fair expectation for a solo builder using Manus: $0 for the first several months, and any four-figure month would require sustained marketing — the exact work the video treats as a footnote. If you want a gentler on-ramp, our roundup of 5 proven ways to make money with AI with no experience leans on methods where distribution isn’t a wall.

One more thing worth knowing about the tool itself. Manus, built by Singapore-based Butterfly Effect, is a serious, well-funded product — Meta announced a roughly $2 billion acquisition in December 2025, as CNBC reported, though the deal was later ordered unwound by Chinese regulators. This is real technology from a real company. The capability is not the issue. The income framing is.

Who this is (and isn’t) for

Manus makes sense if you have a specific, well-scoped task and the budget to feed it — a founder validating an idea fast, an agency prototyping for a client, or someone who already knows how they’ll get users and just wants the build compressed. Have $40–$200/month to spend, a few hours to supervise it, and a distribution plan? You’ll get value.

It’s a poor fit if you’re broke, brand-new, and hoping the tool supplies the customers. It won’t. The credit meter punishes trial-and-error, and the demo’s most impressive-looking output — the finished app — is the cheapest part of building a business.

What to remember

Everything Shark Numbers shows on screen appears to happen: the research runs, the app compiles, the browser clicks. That’s the half that’s true. The half the video skips is the half that pays — installs, retention, and revenue don’t come out of a prompt. U.S. readers should note the FTC has been actively policing exactly this gap: in March 2026 it banned the marketers behind “Air AI” from selling business opportunities over inflated earnings claims, and it has proposed a new Earnings Claim Rule requiring written substantiation for income promises. Treat “idea to launch in an hour” as a claim about building, not earning.

Views and figures are as of the time of review and may have changed since publication.