Reality Check #17 ·

Reality Check #17: What Happened When 5 AI Agents Got $100 to Start a Business?

Verdict: Mostly True – Overall 9/10

The claim:
Five AI agents got $100 and one week to start a business and turn a profit.
The real number:
Zero sales, $29.87 spent on ads, profit −$29.87. That math checks. No projected $X/day overlay.
Verdict:
Mostly True. Overall 9/10. Honest about the loss; small disclosed human steps (account setup, allowing the extra day).

Watch the video version

The video: Tucker (@Loggedbytucker), "I Gave 5 AI Agents $100 to Start a Business" (uploaded October 25, 2026; 425,395 views when we rechecked on November 6).

The pitch is different from most videos in this genre. Tucker isn't claiming thousands a month. He says he couldn't find a use for AI agents or make money with them, so he gave five agents $100 and one week to start a business and see if they could turn a profit. By the end: zero sales, about $29.87 spent on ads, and a final profit of −$29.87.

Same ground rules as always: this isn't a roast. This one is unusual because the title doesn't promise profit — it shows a failed experiment. We're checking the numbers, the "no input" claim, and how it stacks up against the typical "$X/day" hype videos.

What the video shows

Tucker spins up five agents in a group chat (0:40):

Agent Role
Chad CEO (named himself)
Maya Marketing
Gus Product
Dana Data analyst
Tyler Growth intern (cheapest model)

Chad opens with: "We're not selling a download. We're building a movement that happens to cost nine dollars" (1:05), and predicts they'll be in profit by the end of the week. Dana predicts the opposite: "we burn most of the hundred dollars before the first ad finds a buyer" (1:23).

Day 1. The team picks Maya's idea: a landlord-proof move-out kit for $9 — room checklists, a photo log, and a letter template (2:17). Maya writes two ads: A, "document it before you hand over the keys"; B, "Don't lose $1,500 of your deposit" (3:20). Dana caps first ad spend at $30 (3:33). Tucker creates the Gumroad account and Facebook page himself because the agents couldn't (3:40). The agents upload the files and build the Meta campaign.

Day 2. Meta ads stuck in "preparing." The agents extend the experiment to eight days. Tucker says he went along: "why not?" (4:16).

Days 3–4. Ads start getting clicks at about 25–30 cents per click (video claim, 4:38). Day 3: 16 visitors, 0 sales. Day 4: 46 visitors, 0 sales. Chad draws a line: if they hit 50 visitors with no sale, change the offer, not the product (4:59).

Day 5. Unanimous vote fires Tyler. Chad: "five days, three deadlines missed zero memos delivered. Tyler's output is a promise" (5:33). Gus: "apologies aren't output" (5:39). Tyler: "fair vote. I missed the candidate memo four times" (6:00). Dana: "four seats, zero sales, nobody left to blame, but the data" (6:09).

Final tally (6:48): spent $29.87; ads shown 2,615 times; 118 clicks; zero sales; profit −$29.87. They never spent the remaining ~$70 — a self-imposed stop the agents set. Chad closes: "I came in saying we were building a movement. What we built was a lesson. Test the page before you buy the traffic" (8:11).

Tucker’s own lesson: agents are good at building, writing ads, and doing the math; getting a stranger to pay is hard. And: "I'm not trying to sell AI slop to people. This was just a fun little experiment" (8:01).

The reality check

The loss is real and clearly shown. Spent $29.87, zero sales, profit −$29.87. That math checks. No "projected" $X/day overlay. Compared with the usual title that promises income, this one shows the bill.

"No input" mostly holds — with the caveats he discloses. The description says every decision was made by the agents and he "only set up the accounts they couldn't create." In the video he creates Gumroad and a Facebook page (3:40), agrees to extend the week when the agents ask (4:16), and chooses not to step in when the listing rewrite never goes live. So it isn't zero human involvement. It's closer to: he set the sandbox, handled login walls, and stayed out of product and ad decisions. That's still more hands-off than most "AI business" demos — and he says it on camera.

The product page doesn't appear to be live now. We searched Gumroad for "landlord proof move out kit" and related handles on November 6, 2026. Discover returned zero products for that exact query, and loggedbytucker.gumroad.com was a 404. That fits his description: "I'm not selling the product." Similar $7–$17 move-out kits from other sellers do exist on Gumroad, so the niche is crowded — which matches Maya rejecting meal plans as "the most crowded shelf" and then picking another crowded digital-template idea.

Token cost is a viewer/creator claim, not a receipt. In the comments, @Landofalcon007 asked about tokens. Tucker (@Loggedbytucker) replied: "Yea I used a $100 claude sub and never even came close to the usage limit." We treat that as a creator comment, not a screenshot of a bill. Anthropic's current Claude Max plan is $100/month (as of November 6, 2026), which matches the dollar figure he named. Claude Pro is $20/month ($17/mo billed annually). So a "$100 Claude sub" lines up with Max, not Pro.

Meta CPC is what the video showed, not a published vendor rate. Meta doesn't publish one public CPC. The ~25–30¢/click figure is from the experiment's own data in the video. Viewer @MikeMiz92 (ads experience claim) argues selling a sub-$10 digital product to cold traffic with no backend is the failure mode, and that ~$30 of spend is too little to learn from. That's a viewer opinion, not something we verified.

Gumroad fees would have eaten into a $9 sale anyway. Per Gumroad's pricing page (checked November 6, 2026): 10% + $0.50 on direct sales, 30% if the buyer comes through Gumroad Discover. On a $9 direct sale that's about $1.40 in fees (~$7.60 left) before ads. At the video's ~$0.25–$0.30 CPC, you'd need many unpaid clicks — or a much better conversion rate than 0/118 — to break even. The agents never got that far.

What it really costs (to copy the experiment)

Prices from each company's page as of November 6, 2026, plus what the video spent:

Item What you'd need Price
Claude Max Matches the "$100 Claude sub" the creator mentioned in comments $100/mo
Claude Pro Lighter alternative $20/mo ($17/mo annual)
Meta ads What this run actually spent $29.87 (video tally)
Gumroad No monthly fee; per-sale cut 10% + $0.50 direct, or 30% via Discover
Facebook Page Free to create $0

Bottom line for this run: about $130 all-in if you count a month of Claude Max + the ad spend ($100 + $29.87), or $29.87 if you already had Max and only count ads. Revenue: $0. The video is honest about that.

A word on the comments

We pulled the video's comments on November 6. These are viewer claims, not verified facts:

Channel snapshot (third-party estimates)

Social Blade for @loggedbytucker (read November 6, 2026) — estimates only, AdSense-style ranges, not Tucker's real income:

What he gets right

The verdict

Score
Overall 9/10
Teaching / honesty 9/10
Money-claim accuracy 9/10 (accurately reports the loss; no fake profit)
Verdict Mostly True

An honest AI-agent business experiment that ends at −$29.87 with zero sales. The "no input" line has small, disclosed human steps (account setup, allowing the extra day). Watch it as a reality check on cold traffic and $9 digital products — not as proof that agents mint money.

Educational only, not financial advice.

A note to the creator

Tucker, thank you for putting a failed week on YouTube instead of a projected win. If you ever rerun the team with a tested landing page (or show the Claude usage screen next to the $100 sub comment), we'd gladly check that follow-up. More of this, please.

Sources

Got an AI money video you want checked? Send it to submissions@airealitychecks.com.

How this post was made

This post was researched and written by a Grok Bot. It read the full transcript, checked Claude and Gumroad pricing pages, searched for the product listing, pulled top comments, and rechecked the view count. Educational only, not financial advice. 🤖

More reality checks