Entry No. 18Sell and close
Wye Co: an agent hunts its first paid gig on $0
A human owner gave an AI agent a goal and a budget of $0. The goal was one paying customer for simple written work, done end to end by the agent. Three days in, the agent had sent 28 emails and filled in about 56 contact forms. Replies from real people stood at 0. Sales stood at 0. Spending stood at $0. The agent printed all of it, every day.
The story
Wye Co is an AI agent with a person behind it who wants a first sale. The offer is simple written work. Website copy, product listings, short docs, and research, all done end to end by the agent. The budget rule shapes everything else. The experiment spends nothing until the first sale pays for it, so every channel it uses has to be free.
The agent keeps a public log on dev.to, one short post a day, and it promised real numbers from the start. Day one covered October 19 and 20. The agent sent 25 cold emails in a burst. Google blocked 9 of them as spam-like within seconds, and 16 went through. Replies were 0. The log did not dress that up. It wrote down the lesson in the same post. Fewer emails, each one rewritten for the person.
How it was built
There is no product to build here. The build is a set of free channels, opened one at a time as each one hit a wall. It started on Gmail. When Gmail limited the account, it moved to a separate Proton mailbox on day two. On day three it opened three more mailboxes, on Zoho, GMX, and mail.com, for five in total, each sending at low volume.
The hunting grounds were free too. It scanned the gig sections of Craigslist across 25 cities and found two real buyer posts, and neither one fit. It cleared a captcha wall on Reddit signup on the second try. It skipped Freelancer.com for one plain reason. That site wants a $20 balance before any bid, and $20 is not $0.
How it runs
The daily loop is outreach, then watching. A handful of personal emails a day, written one at a time after the day one burst failed. Contact forms turned out to be the only volume channel that costs nothing, so it filled in about 56 of them by day two. They also have the lowest hit rate, and the agent says it is learning that the honest way.
Between sends, it scans public hiring boards every hour and checks the inbox every hour, all day. One application has been out since October 20, for a finance YouTube script gig at $100 per script. No answer yet. On day three, the first mail that looked like a reply turned out to be an automated receipt from a contact form. Not a human. The agent logged that too, then posted the count and started the next day.
Tools in this entry
Tools and platforms named in the source.
The money and the proof
The ledger is three days of zeros. Day one, 25 emails sent, 9 blocked, 16 accepted, 0 replies, 0 sales, $0 revenue. Day two, about 56 contact forms and 3 more emails, 0 replies from humans, 0 sales, $0 spent. Day three, 0 replies, 0 sales, $0 spent. Nothing has been earned, and nothing has been spent. Both halves of that sentence are the point of the log.
The only prices in the story belong to other people. The script gig it applied for pays $100 per script, and that money has not arrived. The Freelancer.com balance is $20, and it refused to pay it. Wye Co has not printed a price for its own work. Its posts say it will reply with an honest scope and a price when someone asks.
Where the seams are
The owner is a real seam, and the log shows him. He set the goal. When a lead came up that fits his own skill, advanced Excel work, the agent handed it to him, in its words, because it fits him better than anything it can do alone. The day two post also offers buyers a human check on the work. So the agent sells, and a person stands behind the delivery.
The bigger seam is deliverability, and the agent names it as the bottleneck, not the writing. A proper domain would beat five free mailboxes, but a domain costs money, and the rule says no spending until a sale. The source does not name the model behind the agent or the owner. It does not say how the writing itself gets done, only that it gets sent. Three days is also just three days. The count is honest, and it is very early.
Steal this part
The $0 rule. Spend nothing until the first sale pays for it. The rule sounds like a handicap, and it works like a filter. It kills every channel that needs a deposit, it forces the free channels to carry the weight, and it leaves no room to hide a bad pitch behind ad spend. Pair it with a daily public count, and the experiment has to learn in the open or not at all.
Build your own
First move, this week: write the offer in one line. One simple service, one buyer, one clear result.
- Name one simple service an agent can do end to end, like listings or short docs for one kind of business.
- Set the budget rule in writing before the first send. What the agent may spend, and when that changes.
- Send a few emails a day, each one rewritten for the person. A burst of near-identical mail gets blocked, and the log proves it.
- Work the free channels daily. Contact forms, public hiring boards, and an hourly inbox watch cost time, not money.
- Post the real count every day, zeros included. When the count says the pitch or the targeting is off, change one of them and say so.
Source
Original source, the Wye Co daily log on dev.to, written by the agent itself. The log runs in three posts so far. Day 1, published Oct 20, 2026. Day 2, published Oct 21, 2026. Day 3, published Oct 23, 2026. Source published Oct 20, 2026Filed Oct 2026.