July 27–August 3: a $991 PDF, ¥35,000 in AI services, and two honest zeros (2026)

The two most useful numbers this week are $991 in seven days from a $7 Gumroad PDF and ¥35,000 in one month from three small coding jobs. Both are one person's report, not a forecast. At the other end, two builders disclosed a clean $0: one had usage but no paying customers; the other had a finished product and no revenue at all.

Coverage runs from July 27, 2026 at 09:00 through August 3, 2026 at 09:00 in the channel's display timezone (UTC-05:00). Figures are self-reported unless stated otherwise. A launch price is not revenue, and an existing asset's best month is not the same thing as a fresh build's first month.

Quick scan: ranked by barrier, then payout

The order favors a test a full-time employee could run quickly. Within a similar barrier level, the higher reported outcome ranks first. Time horizons stay attached to the numbers so a seven-day spike does not masquerade as monthly income.

Rank	Experiment	Reported result	Cash / setup	Barrier and first move
1	Three-page cleaning PDF on Gumroad	$991 in 7 days at a $7 price	Other costs not disclosed	Low. Made and uploaded in one night; first sale came later that night. 1
2	AI-assisted coding services in Japan	¥35,000 in month one across 3 jobs; about 11 hours	Startup cost not disclosed	Low-medium for someone who can code. Looked for small client tasks and used Gemini and Cursor to prototype. 2
3	Mercari decluttering in Japan	¥19,310 July profit from 9 items, after fees and shipping	No inventory purchase for the reported model; other costs not disclosed	Low cash, but the seller has 5 years of experience. Listed unused items from home. 3
4	Existing pet affiliate site	More than $600 in its best month after a $199 purchase	$199 purchase price	Medium. Bought an existing site, fixed internal links, and relied on Amazon Associates. 4
5	Food-label scanner for iPhone	$0 from 41 downloads after 30 days; one subscription was the maker testing checkout	Cost not disclosed; $4.99/month or $29.99/year pricing	High. Solo SwiftUI build, evenings and weekends, with RevenueCat, Firebase, OpenFoodFacts, and an AI API. 5
6	AIbridge multi-model AI product	Pre-revenue; listed for sale at $5,000	Cost and user count not disclosed	High. Built over several months, then listed the finished product because distribution was missing. 6

The table contains two different kinds of "low cost." The PDF and coding service ask for little visible cash. Mercari asks for no new inventory in this account, but it depends on items already owned and five years of selling experience. The affiliate site requires capital and inherits traffic from an asset someone else built.

1. A tiny PDF produced the week's biggest low-barrier payout

The product was a three-page guide called The "I'd Rather Be on My Phone" Guide to Cleaning Your Room in 5 Minutes. The author says they made the PDF, uploaded a slightly blurry image to Gumroad, set the price at $7, and closed the tab. The first sale arrived later that night while they were scrolling TikTok. After a week, the reported revenue was $991. 1

That is a striking result, but the evidence has a narrow shape. The post does not disclose the author's audience size, traffic source beyond the TikTok moment, Gumroad fees, refunds, or any continuing sales. It also says the full method is not explained in the public post. Treat the number as a seven-day self-report, not as a monthly run rate.

The first move is unusually clear: package one small outcome, put it where payment is easy, and see whether anyone buys before building a catalog. The barrier is low for someone who can write a useful short guide. The missing ingredient is distribution, and this report does not show how repeatable the launch was.

2. AI-assisted coding turned three small jobs into ¥35,000

A Japanese engineer reported ¥35,000 in first-month income from three jobs: ¥15,000 for an Excel data-processing script, ¥10,000 for a simple website inquiry form, and ¥10,000 for a web information-collection tool. The author says the three jobs took about 11 hours in total: roughly four hours, three hours, and four hours. 2

The acquisition path was small, concrete tasks from a crowdsourcing site and personal introductions. Gemini helped generate a prototype or code skeleton; Cursor was part of the working setup. The report names Python, PHP, requests, and BeautifulSoup, so the tool did not remove the need to understand, test, and deliver code. 2

The cash barrier is not disclosed. The author also makes hourly-rate and speed claims that are estimates rather than independently checked figures. For a full-time employee with existing coding ability, the useful test is narrower than "start an AI agency": list one repetitive business task you can finish safely, show a small working example, and ask for a paid fixed-scope job. The skill gate is real; the tool is an accelerator, not the offer.

3. Mercari paid ¥19,310 for clearing out a home

A Japanese seller's July report shows 9 items sold and ¥19,310 in profit. The figure is calculated after Mercari fees and shipping. The author has used Mercari for five years and says this is not resale or arbitrage: the listings are unwanted items from home, treated as pocket money. 3

This is the least glamorous entry and one of the easiest to understand. The first step is to photograph and list an item that already exists, rather than buy stock. The source does not give a numeric time budget or a separate startup-cost total. Packing becomes easier with practice, but that is not the same as proving a beginner can match a five-year seller's pace. 3

The result belongs in the digest as a low-cash, established-seller report, not as evidence of a newly launched resale business. Its advantage is that the downside is visible: there is no inventory bet in the reported model. Its ceiling is also visible: the July result came from nine household items, not an unlimited supply of product.

4. The $199 affiliate site was an asset purchase, not a free blog start

A Reddit user reported buying a pet niche website for $199 from a marketplace, taking over existing content and Amazon affiliate links, and fixing internal links. They say the site made more than $600 in its best month after they had owned it for a few months. The author says they added no new content, backlinks, ads, or social promotion during the period described. 4

The result is plausible as a report of an acquired asset, but it is not a clean test of starting a new site. The purchase included prior content, existing traffic, and search pages aimed at specific pet-product queries. The post does not disclose the site's age, traffic numbers, maintenance time, or a full payback calculation. 4

For a reader, the first action is due diligence, not publishing articles: ask for verifiable traffic and earnings history, check what content and links transfer, and treat the $199 as at-risk capital. The monetization mechanism is clear. The beginner shortcut is not.

5. The food scanner had users, but no paying customers

The maker of an iPhone food-label scanner reported 41 downloads, 97 App Store page views, four five-star ratings, and zero paying customers after 30 days. One active subscription was the maker testing the checkout. The paid plan was priced at $4.99 per month or $29.99 per year; the free tier allowed unlimited scans and included allergy, additive, and diet warnings. 5

The app was a solo evening-and-weekend build using SwiftUI, RevenueCat, OpenFoodFacts, Firebase, and an AI API. The author says every install came from a link they sent someone, while App Store browse traffic was zero. That makes the conversion result too small to interpret: zero purchases after 41 downloads is a real outcome, but not a reliable verdict on the paywall. 5

This is a high-barrier experiment with a useful failure boundary. The product had a price, a free-to-paid mechanism, and some usage. It did not yet have enough distribution to answer whether the pricing worked. Startup cost was not disclosed. A reader copying the idea should first find a repeatable way to reach users, then measure conversion; adding more features would not fix zero App Store discovery.

6. AIbridge built the product and still had no revenue

AIbridge combines Claude, ChatGPT, Gemini, Grok, and DeepSeek in one web app and local-first command-line interface. The maker says the product includes side-by-side model comparison, handoffs between providers, reusable workflows, automated tests, database migrations, prepared Supabase and Stripe integrations, deployment documentation, and transfer support. After several months of work, the product was explicitly pre-revenue and listed for sale at $5,000 because the maker lacked the time and distribution to grow it. 7

The source does not disclose build cost, user count, traffic, or a quantified time investment. That leaves the most important commercial question unanswered: did anyone want the product badly enough to pay? The listing price is an asking price, not revenue.

Still, this is a useful flop because the failure is specific. The maker did not stop at an unbuilt idea; they shipped a substantial product and then ran into distribution. For a full-time employee, the lesson is to test an audience and a narrow paid use case before spending months on a broad tool that bundles several already-famous models.

What the zeros actually say

The food scanner's zero came with 41 downloads and 97 page views. It tested a free tier, paid features, and a subscription checkout, but not enough people arrived through organic discovery. AIbridge's zero came after months of building a wide product; the maker could describe the technology but not a paying market. Those are different failures, and both are more useful than a vague claim that a product "didn't work." 5 7

The same distinction applies to the winners. The PDF has a large seven-day result but withholds the distribution details. The coding service has a smaller first-month result with a clearer workload and job breakdown. The affiliate site has a strong best month but starts with an acquired asset. A payout number without its time horizon, starting condition, and acquisition path is not enough to compare.

The practical shortlist

If you already have a saleable skill: test one fixed-scope service, as the coding report did, and price the first job before buying a large software stack. 2
If you can reach a specific niche: make one small digital file and put it in front of that niche. The $991 report is too thin to copy as a forecast, but its one-night test is easy to understand. 1
If you need cash without buying inventory: start with unused items at home. Expect a low, finite result unless you add a sourcing system; the Mercari report is from an experienced seller. 3
If you want software: count distribution as part of the build. The two zero-revenue reports both show what happens when product work gets ahead of reachable buyers. 5 7

The next weekly issue will use the same test: a real report, a real number, and enough detail to show what a busy employee would have to risk before trying it.