midjourney: 11 people, $200 million. outlier or blueprint?
Marco tested an AI CFO that tried to gamble his company's money. It lasted three days. That's how this conversation starts. It ends somewhere very different.
About this episode
Midjourney hit $200 million in revenue with somewhere between 11 and 40 people. No sales department. No marketing team. No free tier. No venture capital. By 2025, revenue was $500 million with 107 people.
Marco thinks it's an unrepeatable outlier — right product, right moment, right platform. Lena thinks it's the blueprint for every company being built today. They debate for 47 minutes. They do not agree.
This is not a structured debate. It's a conversation between two people who respect each other and disagree constantly. There are tangents. There are moments of genuine surprise. Marco tells a story about an AI CFO agent that tried to invest his company's tax reserve in AI ETFs. Lena brings data from the Post AI Index that makes Marco pause. They talk about Valve, Metcalfe's law, Block's 4,000 layoffs, and what happens when coordination becomes free.
What they cover
- The AI CFO that lasted three days (0:00)
- Midjourney's numbers and what they actually mean (5:00)
- The Discord decision: distribution strategy or organizational philosophy? (12:00)
- Why charging from day one is a structural decision, not a pricing decision (19:00)
- No managers: does this scale, and what breaks when it doesn't? (26:00)
- Block, Meta, Intuit: what happens when you discover you're overbuilt (33:00)
- Signals: what Marco and Lena are reading and thinking about (40:00)
- Lightning round: worst AI takes, changed opinions, unspoken worries (45:00)
Transcript excerpt
Marco I tested one of those AI CFO agents last week. The pitch was beautiful. "Replace your finance team with a single prompt." I gave it read-only access to QuickBooks and asked it to categorize three months of transactions and tell me where we were wasting money.
Lena And?
Marco Day one, it categorized our AWS bill as "miscellaneous entertainment." Day two, it flagged the entire engineering payroll as "potential vendor duplication." Day three, I asked it to model our cash runway and it proposed taking our tax reserve and investing it in shares of an AI ETF.
Lena [rindo] No. You're making this up.
Marco I wish. The problem wasn't the AI. The AI was surprisingly good at the mechanical parts. The problem was confidence. It had no sense of when it was guessing. Which, now that I say it out loud, connects directly to what we're supposed to talk about today.
Full transcript coming soon. Subscribe to get notified.
Post AI Sessions
New episodes every other Thursday. Marco and Lena debate one question about the post-AI company.
Get the newsletter →