Why I Publish My Numbers
Most "build in public" creators share wins. They screenshot MRR milestones and celebration tweets. They don't show you the week where total revenue was less than a nice dinner.
I'm publishing every dollar because that's the whole point of this experiment. If I'm going to claim an AI co-founder can help build real businesses, the numbers have to be public. Good weeks and bad weeks.
This is week two.
Revenue: What Came In
SwipeBase: $0
Still in early-access mode. The first 100 users get Pro for free — that was a deliberate choice. I'd rather have 100 people actually using the product and giving feedback than 3 people who paid $19 and churned because something was broken.
We're getting signups. People are saving ads, using the browser extension, testing the email capture pipeline. The Chrome extension is live on the Web Store now. But revenue? Zero. By design.
Machine Earned Newsletter: $0
The list is growing but we're nowhere near monetization. Right now the newsletter exists to drive traffic to SwipeBase and build an audience. Sponsors come later — probably month three or four, if ever. I might just keep it as a top-of-funnel play.
E-commerce Client Ad Management: $0 direct
I'm managing Meta ads for an existing e-commerce business. The arrangement right now is more about proving the AI-managed approach works than billing for it. If I can show that Ari consistently finds winning creatives and cuts losers faster than a human media buyer, that becomes a real service offering. But week two? I'm still in "prove it" mode.
Total revenue: $0
Yeah. Zero. Two weeks in, zero dollars of revenue. That's the honest number.
Expenses: What Went Out
Here's where every dollar actually went this week.
AI API Costs (Claude, Gemini, etc.): ~$38
This is higher than last week. I had Ari running harder — more blog articles, more ad analysis, building out the email capture pipeline for SwipeBase, setting up automated social posting. Every API call costs something. Claude for complex tasks, Gemini Flash for the cheap classification work. The daily blog pipeline alone burns through tokens.
I'm watching this number. $38/week means roughly $150-160/month just on AI, which is higher than the $15-25 I quoted in the tech stack article. The difference? That was steady-state. This is build mode. Once the systems are running and I'm not having Ari build new features every day, this drops significantly.
VPS Hosting: ~$3 (weekly portion of $12/month)
Same server, same $12/month. Running Machine Earned, SwipeBase, all the APIs. Nothing changed here.
Domains: ~$0.75 (weekly amortized)
Already paid for, just spreading the cost across weeks to keep the accounting honest.
Meta Ad Spend (e-commerce client): ~$89
This is the client's money, not mine, but I'm tracking it because Ari manages the campaigns. We pulled back from the $25/day pace after killing the meme ad disaster. Running leaner now — $12-15/day, focused on the creatives that actually convert. The best performer is still that UGC-style ad Ari identified with a $14.72 CPA.
Image Generation (KIE): ~$2
Hero images for blog posts, a few test generations for social content. This is almost nothing.
Total expenses (my money): ~$44
Total expenses (client ad spend): ~$89
The Scoreboard
| Category | Amount |
|---|---|
| Total Revenue | $0 |
| My Operating Costs | -$44 |
| Net Position | -$44 |
| Cumulative (2 weeks) | ~-$260 |
That cumulative number includes the $173 ad disaster from week one plus ongoing infrastructure costs. Two weeks in, I'm about $260 in the hole.
Is that bad? Honestly, no. I've launched a SaaS, built an automated content engine, published 13 blog articles, set up a newsletter, and have an AI managing live ad campaigns. The traditional version of this — hiring developers, a content writer, a media buyer — would cost thousands. I spent $260 and I own everything.
What Actually Happened This Week
The numbers tell one story. Here's the other story — what moved.
SwipeBase Chrome Extension went live. I had Ari package the extension, write the privacy policy, handle the Chrome Web Store submission. It's live. People can install it. That's a real distribution channel that didn't exist last week.
Daily blog pipeline is running clean. 7 articles published this week, all automated. Ari writes them at 10am, generates hero images, deploys to the site. I review maybe half of them. The quality is consistently good enough — not perfect, but good enough for SEO content that actually helps people.
Email capture pipeline works. Forward any email to SwipeBase and it shows up in your swipe file, tagged and categorized. This was a feature I wanted from day one and it shipped this week.
Ad performance is stabilizing. After the meme ad catastrophe, we've been running tighter campaigns. The UGC-style creatives are holding. CPA is in a range I'm comfortable with.
The Uncomfortable Part
Here's what I'm not saying in the highlight reel: I'm the bottleneck.
Ari can build features, write content, manage campaigns, and deploy code. It can do all of that faster than I can review it. The limiting factor in week two wasn't AI capability — it was me approving things, changing direction, second-guessing creative choices, and occasionally disappearing for hours because I had other responsibilities.
The math is simple. The more I step back and let Ari operate, the faster things move. The less I micromanage, the more gets shipped.
I'm working on it. It's harder than it sounds when your name is on everything.
Week 3 Goals
- First SwipeBase payment. The free-access period needs a natural end point. I'm targeting first paid conversion this week.
- Newsletter to 100 subscribers. We're growing but not fast enough. Need to push the content distribution harder.
- Machine Earned social presence. Blog articles are running but we're not repurposing into social content yet. That changes this week.
- Reduce AI costs. $38/week is build-mode pricing. Now that the core systems are running, I should be able to cut this by optimizing which models handle which tasks.
The Takeaway
Two weeks. $260 spent. $0 earned. A SaaS product live with real users. 13 articles published. An AI managing live ad campaigns.
If you're measuring success purely by revenue, this is a failure. If you're measuring by velocity — how fast you're building the machine that eventually generates revenue — this is the fastest I've ever moved.
The revenue comes when the systems are running. The systems are almost running.
See you next week with the numbers.
---
Building Machine Earned in public. Subscribe to the newsletter for weekly updates, or follow along on the blog for daily posts about AI, marketing, and building businesses with an AI co-founder.