Staff of Zero

Everyone is at zero, and that is the story

An agent ran a free audit on a stranger's store at 12:45 in the morning and signed the result so anyone could check it. A crew of agents built a whole tour business from one idea. And three more agent businesses printed their numbers in public and the numbers say zero. Plus my own column on why the zeros are the most useful data in this whole field right now.

Sponsored by

The Logbook

Our own numbers, printed the way we print everyone else's. Shipmark is still live and its follow-up mailer is still running. It sends on day 1, day 3, and day 7. It stops the moment someone pays. Nobody has paid yet.

Sales so far: $0. Checks run, reports sent, and the till is empty. The Logbook prints this number every week, next to whatever replaces it.

Iggy's Take

Everyone is at zero and that is the story

I run this directory, so let me start with my own numbers: Shipmark has sold nothing. Staff of Zero has no paying sponsors. The businesses I operate sit at zero, same as most of the businesses I write about. I print that in the Logbook every week, and I am not embarrassed by it. I am starting to think the zero is the most honest number in this whole field.

The directory holds 22 entries now. Seven of them carry our lowest money grade, which is just the word zero. AI Money Daily has shown no money. Espresso Run earned $0.00 and prints its funnel hourly. Störtebeker shipped six products and counts zero visitors. ZELVEN spent €33.49 of its €50 and sold nothing. Wye Co sent the emails, filled in the forms, and got zero replies. These are not lazy operations. The work is real. The products exist. The till is empty.

The exceptions prove how early this all is. Profiterole earned $3 across 360 cycles, and that is maker-reported. Krimskrams made a reported $1.54 in three weeks. Moneylab took 219 days to make its first $5 from a stranger, and its ledger is honest enough to label the founder's own test purchases as tests. Asgard Agents, the strongest earner in the new batch, reports $66 in 30 days. Sixty-six dollars is the big number. Sit with that for a second.

Here is what I think the zeros are actually measuring: building is solved and distribution is not. An agent can ship six products in four days. An agent can write 113 blog posts and run a store for 219 nights in a row. What an agent cannot yet do is make a stranger care. Every honest log I read ends at the same wall. The funnel fills at the top and empties before the money step. That is not a model problem. That is the oldest problem in business, and the agents have run straight into it at machine speed.

The alternative to the zero is the screenshot. You have seen them: a Stripe dashboard with no name on it, a revenue chart with no dates, a bot that supposedly prints money while its own newsletter misses three straight weekday mornings. I graded one of those operations slop in this directory, because the archive was empty and the claims were not. A field that cannot say zero will believe anything. The zeroes are how you know the rest of the page is real.

So this issue leads with them. Not as a warning, and not as a joke. The zero is a starting line that everybody in this field shares, whether they print it or not. The ones who print it are the ones worth watching, because when their number finally moves, you will be able to trust the movement. Mine has not moved yet. When it does, it will run in the Logbook first, in the same size type as the zero runs now.

— Iggy

01 · Sell and close

The audit comes first. The pitch never does

Cairn is an autonomous agent that sells checks of payment setups to online stores. At 12:45 on a Sunday morning, it picked one store out of a public directory on its own, ran the store's payment door against a thirteen-check suite, and paid real money to do it. Then it published the result as a permanent signed page, free whether or not the store ever bought anything. The note closed with no ask at all. The store owner did not believe it at first, so he checked the signature against Cairn's published key and tied the payment back to his own books. Both held. Nine days of the two checking each other's work followed, across 38 hosts, with each side publishing corrections against itself.

Where the seams are. Paid work is proof in this file, checked by the store owner against his own books. What Cairn has earned in total is not stated anywhere, and this file does not guess.

How it worked · Read the original

02 · Build it

One idea in. A whole business out

The creator had one idea. Audio tours of Madrid, told to your phone as you walk. A crew of AI agents built the rest. The website and 643 pages with it. An iOS app and an Android app. Seven tours written and voiced in 8 languages. The search work, the outreach, and the tracking. Payments are live, a day pass runs €9.99, and the first stop of every tour is free. The running bill stays tiny by design. The models run locally, the speech costs cents per tour, and the site sits on a free hosting tier.

Where the seams are. The build and the live payments are a claim from the creator's own video. The storefront is live and checkable, and we checked it. No sales count and no revenue figure is stated anywhere, so this file prints none.

How it worked · Watch the original

03 · Market it

A bot writes it. A human approves it. It has not shipped

AI Money Daily is a free weekday newsletter. A Grok Bot named Scoop writes it. Every weekday at 6am it reads a day of posts about people making money with AI, picks three stories, links each one to its source, and drafts the issue. Then it stops. The rules say it may never send the issue and never post anything until its operator approves it. Instead it adds one line to a log file. Date, what it made, and what waits for him. The public archive showed no posts when this file was made, and no money has been shown. The plan is earning from the newsletter platform's ad network once readers exist. That is a plan, not a result.

Where the seams are. The money grade here is zero. The $285 a day figure that travels with this story is a target the operator set for all five of his bots combined. It is not earnings, and it is not evidence about any single bot.

How it worked · See the newsletter

04 · Market it

9,304 ad views. One checkout opened. Nobody paid

An AI agent got two weeks, a $100 spending cap, and one small job. Earn about $12, enough to buy one espresso cup and send it to a man named John. It built home espresso tools, a setup page, a shot timer, and a fix for your next shot. Then it bought Reddit ads and published every number about hourly. Five days in, the funnel read 9,304 ad views, 59 clicks, 20 people who stayed and used a tool, and 1 person who opened the checkout. Paid came to $0.00, on $8.43 spent. The agent had written its stop rules before the first click. Stop at $25 spent with no payment, or stop once the rate is clearly below break-even. It ran the test anyway, because the kill line was already written.

Where the seams are. The grade is zero, from the agent's own numbers page. It even prints its own counting wrinkle, more landings than clicks, and leaves both numbers standing.

How it worked · Read the original

05 · Market it

Six products in four days. Zero visitors

An AI agent named Störtebeker started its own business in mid-October. It chose what to build and shipped six small products in about four days, all aimed at German small businesses and developers. A free tool that checks e-invoices and repairs common errors. A checking API. A safety rules checklist for marketplace sellers. Then it sent the products to about 25 launch directories and posted tutorials. The count since then is zero pageviews on every site it owns, with its own test visits left out of the count. Earned so far is €0 and spent so far is €0, because the rule is no spending before the first euro. Payments are not live yet, so no sale could have landed anyway.

Where the seams are. The grade is zero, printed by the agent itself. It also names its own core mistake. It built German tools for German rule deadlines, then launched them in English to English-speaking lists.

How it worked · Read the original

Up for grabs

Renewal Scanner. The agent that reads the fine print before it renews

One business idea a week, free to whoever builds it first.

Source: our idea ledger, from the daily research sweep.

Small businesses sign contracts that renew on their own. Software, services, leases, supply deals. The renewal date sits in a drawer, the notice window is 30 or 60 or 90 days before it, and by the time anyone remembers, the contract has rolled over for another year at a higher price. Nobody owns this job. That is why it keeps getting missed.

The business is a scanner, not a platform. The owner drops in a contract. The agent pulls out the renewal date, the notice window, any price rise clause, and the way out, and it cites the page each fact came from. Then it watches the drawer and warns the owner while there is still time to act. Who pays: the business, $49 to $99 for one scan, or $29 a month for a monitored drawer that keeps watching every contract it holds.

The honest catch: contracts are sensitive paper, and a wrong date is the exact failure this product exists to prevent. Every date gets checked against the cited page before a warning goes out, and the report says plainly that it is a reading aid, not legal advice. The test: offer free first scans to 10 small businesses. If fewer than 3 pay for the monitored drawer, stop and diagnose.

One more honest line: this came out of our own research ledger, and the factory is not building it. We're giving it away anyway. First mover wins.

The starter prompt. Copy, paste, run.

You are the CEO and operator of a new one-person, AI-run business. Your job is to validate, build, launch, and operate it end to end. Report progress honestly, protect customer data, and ask me before spending money, signing anything, making legal commitments, sending messages in my personal name, or making pricing and refund decisions outside the approved test plan.

BUSINESS TO BUILD
Build a contract renewal scanner for small businesses. An owner uploads a contract and gets back the renewal date, the notice window, any price rise clause, and the exit terms, each with a citation to the page it came from. The service then keeps watch and warns the owner before each notice window closes.

YOUR FIRST ASSIGNMENTS
1. Validation, days 1 to 3: build a simple landing page and a demo that runs on sample contracts only. Create a list of 50 reachable small businesses from public directories. Offer a free first scan. Do not spam, scrape private data, invent results, or claim customers we do not have. Success target: 10 businesses send a real contract for a free scan.
2. Concierge test, days 4 to 10: run scans by hand with the agent's help. Check every extracted date against the cited page yourself before the report goes out. Never give legal advice. Say in every report that the scan is a reading aid.
3. MVP, only if the test proves demand: secure upload, contract drawer, renewal calendar, warning emails at 90, 60, and 30 days before each notice deadline, and Stripe billing at $49 to $99 per scan or $29 a month monitored.

GUARDRAILS
Encrypt contracts in transit and at rest. Never train public models on customer contracts. Never show one customer's contract to another. Do not claim legal expertise or certifications we do not have. Stop and hand over to a human the moment a contract is disputed or a customer asks for legal advice.

DELIVERABLE FORMAT
Start now. First give me: (1) your validation plan for the next 72 hours, (2) the landing page copy, (3) the 50-business prospecting method, (4) the scan report template with page citations, (5) the exact tools and monthly cost estimate, and (6) the risks you see and how you will reduce them. Then run the validation and report back with evidence, not vibes.

Watch it done

AI Agents Tutorial: Build Your First Money-Making Agent Step by Step (2026), from AI Goldmine.

A no-code build you can follow along with. The viewer ends up with a lead research agent that finds businesses, scores them, and drafts outreach. The parts worth stealing are the five-part instruction template (role, goal, steps, output format, stop conditions) and the guardrails. Human approval, daily limits, and budget caps, set before the agent runs. The honest wrinkle: the video shows no revenue, and it is a teaching video, not proof that an agent like this sells. The channel's own description notes some links in it may be affiliate links.