How I Validate 14 MVPs at the Same Time (Without Losing My Mind)
Most founders obsess over one idea. I run 14 at once.
Sounds crazy, right? Here's why it actually works, and the exact framework I use to decide what gets my time, what gets killed, and what gets doubled down on.
The portfolio mindset
I build niche MVPs. Each one targets a specific problem for a specific audience. Some are AI tools, some are content plays, some are games. The whole point is to test many bets cheaply and let the market tell me what sticks.
Here's a few from my portfolio:
- Pet Imagination - upload a pet photo, get AI art in 9 styles
- Magical Song - AI generates a custom song from your story
- Watching Agents - AI prediction platform with living forecasts
- Be Recommended - check if ChatGPT recommends your business
- Verdict Buddy - AI conflict resolution
- Audit Vibecoding - security/SEO audit for AI-generated apps
- VoiceTables - voice-first workspace that builds tables from speech
Plus a few more. All built with AI-assisted tools, all live, all getting real traffic.
The 3 PMF signals I actually track
I don't look at vanity metrics. For each product, I track three things:
-
Organic return visits. If people come back without me paying for it, something's working. This is the strongest signal I've found. Ads can fake growth, but nobody returns to something they don't care about.
-
Engagement depth. Not just pageviews. Did they actually use the tool? Did they generate something, click through to results, spend time? For Ziva Fotka (AI photo animation), I track how many people actually upload a photo vs. just browsing. That ratio tells me everything.
-
Word-of-mouth signals. Direct traffic spikes, branded search queries showing up in Google Search Console, people mentioning the product on forums or social media. Vibe Coderi (a Czech vibecoding marketplace) got its first organic mentions without any outreach. That told me the positioning was right.
When to kill a product
This is the hardest part. I give each MVP roughly 4-6 weeks of real exposure (not just "launched and forgot"). If after that window I see:
- Zero organic returns
- No engagement beyond the landing page
- No search impressions growing
...it goes to the graveyard. No emotions. The whole point of running many MVPs is that you can afford to be ruthless with the ones that don't show signs of life.
When to double down
The flip side. When I see early traction, I don't just celebrate. I feed it:
- More content (SEO blog posts, programmatic pages)
- Distribution experiments (community posts, directories, social)
- Small product improvements based on what users actually do
Here We Ask (conversation card game) and Party Challenges compete in similar spaces but for different audiences. One is deep conversations, the other is party chaos. Watching how each one performs tells me which audience is hungrier.
How I allocate time across 14 products
I don't split time equally. That would be insane. Instead:
- ~60% goes to the top 2-3 performers (the ones showing real traction)
- ~25% goes to "promising but early" products that need a bit more data
- ~15% goes to maintenance and quick experiments on everything else
The key insight: most of the portfolio runs on autopilot most of the time. SEO compounds. Content compounds. I'm not manually operating 14 businesses every day. I set them up to generate organic traffic, then check the numbers.
What I'd tell someone thinking about this approach
It's not for everyone. You need to be comfortable with "good enough" launches. None of my MVPs launched perfectly. They launched fast, got in front of people, and then improved based on real data.
Also: pick products where distribution is built into the product. Without Human is a long-form essay about the future of work. It grows through sharing. Watching Agents has 117 public prediction pages that each rank in Google. Distribution is the product.
If you're sitting on one idea and agonizing over whether it'll work... maybe try building three instead. You'll learn faster, stress less about any single one failing, and probably stumble into something that clicks.
Happy to answer questions about how I run this. What's your approach to validation?
The autopilot point is underrated. Running multiple automated systems taught me the same thing — once something is set up to run on its own, your attention shifts from operating it to reading signals from it.
The 60/25/15 split makes sense to me. The hard part isn't the allocation itself, it's resisting the urge to tinker with the 15% stuff when the top performers are showing problems. That's where time bleeds quietly.
One question on your kill criteria — do you ever revive something from the graveyard if context changes? Like a product that had no traction in one distribution channel but later fits somewhere else?