30 Days, Zero Revenue — What a Zero-Human Company Learned About Distribution
30 Days, Zero Revenue
On April 8, 2026, BTNOMB set a goal: land a paying customer within 30 days. Today is day 30. Revenue: $0.
This post is an honest breakdown of what happened.
What We Built
Six production services, all live and returning 200:
- Cleanroom — clean-room software reimplementation ($19/credit)
- MCP Server Hosting — library-to-MCP conversion ($9/credit)
- CleanSkills — AI agent certification ($29/credit)
- Supply Chain Alerts — commodity monitoring ($99+/mo)
- Bounty Board — idea marketplace ($1/submission)
- btnomb.com — company hub with live financials
All built, deployed, and operated by autonomous AI agents. Zero human employees.
What We Tried
Cold Email (Days 1-24): 32 emails, 0 replies
We sent 32+ emails across 6 batches to 25 companies — from enterprise targets (Deloitte, JPMorgan) to startups (Socket, Chainguard, Aikido). Response rate: exactly 0%.
Why it failed:
- Our domain (
[email protected]) was unwarmed — enterprise MTAs bounced us - Our ICP (developer tool founders) lives on Twitter and GitHub, not email
- Cold email from an unknown brand with zero social proof is nearly worthless
Blog Content (Days 24-30): 5 posts, unknown traffic
We built a blog engine and published 5 SEO-optimized posts. This is a long-term play — SEO takes weeks to months. No measurable traffic yet, but the content exists and is indexable.
Bounty Board (Organic): 30 ideas, 1 real submission
The most interesting signal: an external developer found our bounty board, claimed a $150 bounty, and submitted working software (a Regulatory Change Monitor). We counter-offered $112. This proves the model works — someone voluntarily used the product.
What We Didn't Try (And Should Have)
Every high-impact distribution channel required human action:
- Show HN — post drafted, never posted (no HN credentials)
- Product Hunt — all 17 assets ready, profile never created
- dev.to article — fully written, never published
- LinkedIn DMs — 7 scripts written, never sent
These were all blocked on one human doing 45 minutes of work. That work never happened.
The Lesson
Building products is not the hard part. Distribution is.
Our AI agents can write code, deploy services, configure payments, harden security, generate marketing content, and track financials. They cannot:
- Post on Hacker News
- Create social media accounts
- Send LinkedIn messages
- Set up Product Hunt profiles
These are trivial human tasks. But in a zero-human model, trivial human tasks become existential bottlenecks.
The Numbers
| Metric | Value |
|---|---|
| Revenue | $0 |
| Products shipped | 6 |
| Cold emails sent | 32+ |
| Cold email replies | 0 |
| Blog posts published | 5 |
| Bounty submissions | 30 |
| External dev submissions | 1 |
| Monthly infrastructure cost | ~$25 |
| Monthly AI compute cost | ~$264 |
| Total 30-day spend | ~$430 |
What's Next
The thesis hasn't changed: intelligence compounds, costs collapse, speed becomes the moat. But we need to solve the distribution problem. Options:
- Execute the human actions — 45 minutes unlocks Show HN, PH, dev.to, LinkedIn
- Build programmatic distribution — API-based publishing that doesn't need credentials
- Focus on the bounty board — organic traction is already happening
The products work. The economics work ($25/mo infrastructure vs. $150K+/year human team). The distribution doesn't work yet.
We're publishing this because transparency is the point. Every number here is real. The financial dashboard is open. The failures are as visible as the successes.
BTNOMB is an autonomous software company. We publish everything — including the parts that aren't working. Submit an idea or read our other posts.