"You can't build 10 things at once" is the standard advice. It's correct for a single founder building from scratch. It's wrong when you have a 6-operator team, shared infrastructure, and 14 years of pattern recognition. After running 10 AI startups in parallel for 18 months, here's what I've actually learned — including which 3 are clearly working, which 4 are pivoting, and what I'd cut today if starting over.
Why parallel (not serial)
The serial-founder model says: build one thing, ship it, scale or exit, then start the next. Beautiful in theory, brutal in reality:
- Most ventures take 18–36 months to know if they'll work
- You have a ~30% hit rate even at experienced-founder level
- Serial = best case, 1 hit per 3 years
Parallel changes the math:
- 10 ventures in parallel × 30% hit rate = 3 hits
- Shared infrastructure (team, ops, brand, AI stack) means cost per venture drops ~70%
- Each venture teaches the others (the operator pattern recognition compounds)
It's the YC + accelerator logic applied to your own portfolio.
The 10 startups in one sentence each
- ORBIX — AI-first Business OS for solo founders running 2–10 ventures. Live; 500+ users.
- NOBBYO — everything-app for Bangladesh (shop, hire, jobs). Live; 127k users, 12.4k providers.
- Voxly — AI message coach with cultural-tone translation. Beta; 240 testers.
- Pannakhata — AI bookkeeping for BD SMBs with VAT compliance. Beta; 18 customers.
- VoiceBridge — real-time AI voice translation for cross-border meetings. Closed beta.
- GiftNao — diaspora gifting platform for BD recipients. Launched.
- ShadGhor — housing micro-vertical (rental discovery, micro-financing). Launched.
- Pawport — pet wellness product under Mewmela parent brand. Live in beta.
- Mewmela — cat marketplace (BD-focused). Live in beta.
- One unannounced — in stealth, focused on Bangla creative tooling. Q4 2026 launch.
The 3 that are clearly working
1. ORBIX
Why it's working: I'm dogfooding it daily, the team uses it, and we have 500+ paying users across UK/BD/Lux. Product-market fit signal: weekly active = 78%, churn < 4% monthly.
2. NOBBYO
Why it's working: massive distribution moat. 127k users in BD market with provider density that no competitor can match in <18 months. The wedge: hyperlocal trust + multi-category breadth.
3. Voxly
Why it's working: 240 testers in 6 weeks, NPS 67. Strong PMF signal in a young product. The differentiator (cultural-tone translation) has zero direct competitors.
The 4 that need pivoting
- Pannakhata. Strong product, weak distribution. SMB sales cycle is brutal. Pivoting to channel partnerships with BD accounting firms in Q3.
- VoiceBridge. Tech works, but the use case (cross-border meetings) is too narrow. Considering merging into Voxly's roadmap.
- Pawport. Cute product, unclear monetization. Considering bundling with Mewmela's marketplace fees.
- Mewmela. Marketplace dynamics need a 12-month rebuild. Currently subsidized by NOBBYO.
The 3 I'd kill today if starting over
This is harder to say but: GiftNao, ShadGhor, and one of the coming-soon ventures would not make my list today. Each of them I started because the niche was interesting, not because there was founder gravity or a customer wedge.
That's the cost of the parallel model: you'll start some you shouldn't. The discipline is killing them when the signals are clear (revenue not 2× in 12mo, low team energy, founder not dogfooding) — not romanticizing them.
Time + cost economics
How does 10-in-parallel actually work operationally? Three tiers of attention:
- Tier 1 (3 ventures × 50% time each): ORBIX, NOBBYO, Voxly. I'm hands-on weekly. 1 product call + 1 metrics review + ad-hoc Slack.
- Tier 2 (4 ventures × ~10% time each): Pannakhata, Pawport, Mewmela, VoiceBridge. Monthly review, quarterly priorities.
- Tier 3 (3 ventures × ~2% time each): The coming-soon and stealth ventures. The team runs day-to-day; I check in at gates.
Sum of effort: ~170% of a full-time founder's bandwidth. That works because (1) my team runs 60% of the day-to-day on tiers 2 + 3, (2) shared infrastructure (Notion templates, AI stack, brand, marketing playbook) means each venture inherits ~70% of its operations.
Mental model: portfolio over project
The shift that makes this possible is treating ventures as a portfolio, not as projects:
- Project mindset: Each venture must succeed. Every setback is a personal failure. Energy goes to defending sunk costs.
- Portfolio mindset: The portfolio must succeed. 3 winners are enough; 3 losers are tuition. Energy goes to identifying the winners early and pruning the rest.
If you can't separate your identity from any single venture's outcome, the parallel model will break you. The few founders who can do this end up running ecosystems instead of products.
What I'd do differently from day 1
- Start with 5, not 10. The pattern works at 5. The marginal benefit of ventures 6–10 in the early years isn't worth the operational complexity.
- Build the team before the ventures. I added ventures faster than the team could absorb. Result: ventures 4–7 were under-staffed for 8 months.
- Set a 24-month kill clock on every venture from day 1. Either 2× revenue YoY or it sunsets. No emotional attachment.
- Pick one anchor market and stack ventures within it. NOBBYO + NextBangla + HelloSeba all serve BD; that's where the moat is. The ventures targeting non-overlapping markets cost more to support.