Issue 1 · 2026-W43 · 19-25 October 2026
Five AI monetization plays with a public source and a number
The plays that cleared this week's scan: a public source, a figure somebody stated, and a way in.
-
$9.99/month for a todo app with email that was generated with AI
A task manager built by one person with heavy AI help, sold at $9.99/month after a two-week trial. Its hook is email: incoming mail becomes tasks.
- Price
- $9.99/month after a two-week trial
- Built by
- one person, AI-assisted: research, UI flows, schema rewrites, performance tuning
- Demand. A defended price, not a market: the builder says ads could not cover the AI bill, and nobody publishes users or MRR.
- Why it could work. Email-to-task is a daily habit, and a recurring model bill gives a subscription an honest job.
- What makes it buildable. Inbound mail parsed into tasks, a forwarding address or inbox OAuth, Stripe trials, and cost control per user.
- Who is already there. Todoist, TickTick, Things and OmniFocus sell the same habit with years of polish; a weekend clone is the nearer threat.
- What kills it. Trial-to-paid is the whole business. If a user's model calls cost more than $9.99 a month, growth makes the loss bigger.
- Cost to test. Two to three weeks and about $0 in cash; the first subscriber who renews past month two.
-
Productize spreadsheet cleanup at $1,000/mo per client: the work is formatting, not analysis
A productized service: spreadsheet cleanup and repair on a monthly retainer at $1,000 per client. The work is formatting and reconstruction, not analysis.
- Price
- $1,000/month per client
- The work
- repairing and rebuilding workbooks, formatting rather than analysis
- Demand. A price and a job type that recurs, not a named payer: the post shows no invoice, so check one real quote.
- Why it could work. Every operator has a workbook only one person understands; one client meets the title, the second is margin.
- What makes it buildable. LLMs to write and repair formulas, Power Query and VBA, Excel Copilot, an intake form, a shared drive.
- Who is already there. Upwork and Fiverr quote hourly for the same repair, and Excel Copilot is the free first draft.
- What kills it. The sheet gets fixed and the retainer feels like a one-off; formatting invites a cheap hourly comparison.
- Cost to test. One week of evenings and $0: quote ten people. Two replies at $1,000 is the signal.
-
Grow an audience by replying, then sell sponsored posts while a small SaaS compounds
An audience-first play: reply your way from 0 to 13k followers in 200 days, then monetize the attention while a small SaaS compounds underneath.
- Receipt
- $1,044 MRR on a public TrustMRR page
- Took
- 200 days, 0 to 13k followers
- Demand. The revenue is a paying-subscriber figure tied to a page someone else can audit; the followers are distribution, not demand.
- Why it could work. Replies are the cheapest reach there is, and the following compounds while the SaaS collects.
- What makes it buildable. X search and notifications, a drafting tool, a log of what got replies, TrustMRR for public revenue.
- Who is already there. Half of build-in-public X runs this play; newsletters own attention with less platform risk.
- What kills it. 200 unpaid days before $1,044, and X owns the algorithm, the API and the reach.
- Cost to test. Zero cash, 200 days of replies; give it 30 days and watch whether followers move at all.
-
Generate a digital product with ChatGPT, list it on Gumroad: $103.45 in the first 24 hours
Draft one digital product with ChatGPT, list it on Gumroad, and let the first day's sales price the idea. The author reports $103 in 24 hours.
- Takes
- one day from draft to listing
- Reported
- $103 in the first 24 hours
- Demand. A day of takings is a payer, not a projection: small, but it cleared and the product delivered itself.
- Why it could work. The loop closes in a day, so ten niches cost ten days and each one teaches a title and a price.
- What makes it buildable. An LLM for the file, Gumroad for checkout and delivery, one post where the audience already is.
- Who is already there. Gumroad is crowded with prompt packs and templates; a buyer with the same model can regenerate the file.
- What kills it. Most listings sell zero; if the $103 was a launch-day spike, day two tells you.
- Cost to test. One day and a listing, cash near zero; nothing sold in 48 hours means fix the offer.
-
Sell AI services on Fiverr from a low-cost country, where $132.80 is the honest starting figure
Sell AI services as Fiverr gigs. The author did it from Pakistan and reports $132 in total, the honest starting figure rather than a run rate.
- Reported
- $132 in total, orders paid through Fiverr
- Delivery
- prompt work: copy, edits, small automations
- Demand. Marketplace demand, not yours: buyers released the money, but Fiverr put the gig in front of them.
- Why it could work. No audience needed because the marketplace matches buyers, and $132 goes further where the author lives.
- What makes it buildable. Frontier models for the delivery, Fiverr's ranking for distribution, three samples, replies under an hour.
- Who is already there. Thousands of AI gigs undercut each other, plus agencies bidding the same brief.
- What kills it. Fiverr owns the customer, the ranking and the account; prices floor out and reviews take months.
- Cost to test. Two weeks, no cash: one gig, three samples, hour-fast replies. Did anyone order in month one?
The six-part brief is the paid row
Each pick above is the short version: what it is, the facts the source states, and a line each on demand, why it could work, what makes it buildable, who is already there, and what kills it. The paid index adds the cost to test, the competition in detail, and the risks priced.
Get the next issue by email
Free, weekly, five plays, one click to unsubscribe.
The address is used for this letter only.