Appearance
Chapter 4 — Solo SaaS $1M
💰
"14,000 lines of raw PHP mixed with inline HTML, CSS in style tags, JS in script tags.0 employees. $132K MRR. $1.65M ARR." — Pieter Levels (@levelsio), tweet 4.8M views
You'll learn
- How to build solo SaaS on top of AI gen APIs (Flux, Replicate, Fal)
- Why "ship before perfect" + build-in-public = competitive moat
- 3 cases from $0 → $1M-$401M ARR
- Underserved AI gen niches
- Tax + payment stack for global founders
01 Pieter Levels — solo indie hacker $1.65M ARR
Profile
| Item | Number |
|---|---|
| Name | Pieter Levels (@levelsio) |
| Home | Netherlands, lives Bali / Netherlands |
| Background | 10 years "build in public", 70+ failed startups |
| X followers | ~600K |
| Employees | 0 |
Portfolio (May 2026)
| Product | Status | ARR/MRR |
|---|---|---|
| PhotoAI | AI headshot/portrait | $1.65M ARR ($132K MRR Nov 2025) |
| InteriorAI | AI interior design | ~$300K ARR |
| NomadList | Remote community | ~$500K ARR |
| RemoteOK | Remote job board | ~$300K ARR |
| fly.pieter.com | Browser flight game (vibe-coded) | $0 → $1M ARR in 17 days |
| Total | ~$3M ARR |
fly.pieter.com — case study
| Milestone | Number |
|---|---|
| Launch | Mar 2025 |
| 17 days later | $1M ARR ($87K MRR) |
| Build-in-public tweet | 4.8M views |
| Stack | three.js + raw JS + Cursor |
| Build time | 3 days |
| Team | 1 (Pieter) |
Defining quote
"Without AI it would have taken me 10-100x more time. AI really is a creativity and speed maximizer.""Ship fast. Ship ugly. Ship in public." — Pieter Levels
02 Maor Shlomo — Base44, $80M exit / 6 months
| Item | Number |
|---|---|
| Founder | Maor Shlomo (Israel, solo) |
| Product | Base44 — no-code AI app builder |
| Launch | Feb 2025 |
| Users in 4 weeks | 250K |
| Month 1 revenue | ~$1.5M |
| Exit | $80M acquisition in 6 months |
Lesson: Solo founder + AI app builder + execution velocity = new 2025-2026 pattern.
03 Matthew Gallagher — Medvi, $401M year 1
| Item | Number |
|---|---|
| Product | Medvi (healthcare-adjacent) |
| Setup cost | $20K + 2 months |
| Year 1 revenue | $401M 🤯 |
| Projected | $1.8B/year |
| Engineers hired | 0 (only contractor for custom parts) |
Stack:
- ChatGPT / Claude / Grok — code
- Midjourney + Runway — ad creative
- ElevenLabs + custom agent — customer service
First case approaching $1B from solo founder. Sam Altman predicted this. Medvi delivered it.
Note: $401M ≠ profit. Top-line revenue. Still impressive for solo.
04 Pattern across all 3 cases
5 recurring factors
1. Wrap 1 good API + specific niche
- Pieter: Flux + portraits
- Maor: LLM + app builder
- Matthew: multiple LLMs + healthcare niche
2. Ship early, ship ugly
- PhotoAI = 14K lines of "raw" PHP
- Base44 first version "good enough"
- No over-engineering
3. Build in public
- Pieter: 600K X followers over 10 years
- Maor: weekly progress shares
- Audience before product
4. Underserved niche
- AI headshot (Pieter) — corporate needs photos, no pro service
- No-code AI builder (Maor) — non-devs want to build apps
- Healthcare booking (Matthew) — fragmented market
5. Extremely lean cost structure
- 0 employees
- $50-500/month cloud
- $0-200/month tools
05 Build pipeline — 6 steps
Step 1. Pick niche
3-question framework:
- What pain do you experience repeatedly? (e.g., visa photos)
- Who's the audience? (e.g., people emigrating)
- Which AI gen tool solves it? (Flux + LoRA)
Step 2. Validate in 24 hours
- Create landing page (Webflow / Framer / Carrd) — 2 hours
- Product description + waitlist signup
- Post Twitter / Reddit / Facebook group
- Measure: 24-hour signups?
Target: >20 signups in 24 hours = viable niche.
Step 3. MVP in 7 days
Simple stack:
- Frontend: Next.js / SvelteKit / raw PHP (Levels-style)
- Auth: Clerk (free tier)
- DB: Supabase / Postgres
- AI API: Replicate / Fal.ai / OpenAI
- Payment: Stripe
- Deploy: Vercel / Cloudflare Pages
Step 4. Pricing
| Model | When to use |
|---|---|
| Pay-per-use | Generation product ($0.05-0.20/image) |
| Subscription | Recurring ($9-49/month) |
| One-time | Lifetime deal launch ($29-99) |
| Hybrid | Sub + credit top-up |
PhotoAI: $9-39/month + credits. Base44: $29-99/month + usage.
Step 5. Launch sequence
| Day | Action |
|---|---|
| -7 | Build waitlist tweet thread |
| 0 | Launch ProductHunt + Hacker News + Reddit + Twitter |
| 1-7 | Daily build-in-public update (signups, MRR) |
| 14 | Post-launch iteration based on feedback |
| 30 | Measure $MRR + retention |
Step 6. Iterate
- Weekly: pricing test
- Bi-weekly: feature ship
- Monthly: review churn + add referral
06 20 specific niches (May 2026)
Concrete niches
Image gen
- AI executive headshots
- AI passport/visa photos
- AI wedding photos (cultural backdrop)
- AI product photos for TikTok Shop/Shopee sellers
- AI interior design
- AI logo + brand identity
- AI CV/LinkedIn profile photos
Video gen 8. AI TVC mockups for SME pitches 9. AI UGC ads for TikTok Shop 10. AI karaoke MV 11. AI explainer videos for B2B SaaS 12. AI mini-series in local language
Audio 13. AI dubbing local language for foreign videos 14. Genre-specific AI music production 15. AI audio books in local language 16. AI voice clone for podcasts (own voice)
3D 17. AI 3D assets for Roblox creators 18. AI 3D models for e-commerce (360 view)
Multimodal 19. AI full brand identity pack (logo + ad + voice) 20. AI yearbook / portfolio for students
07 Prompt pack
5 templates
1. Niche validation (Claude)
I want to build an AI SaaS in [TOPIC] niche.
Analyze:
1. TAM global vs [local market]
2. Top 3 competitors + price + USP
3. 5 user pain points
4. 3 differentiating marketing angles
5. Pricing model + suggested price2. Landing page copy (Claude)
Write landing copy for [product name]:
- Headline (1 line, <12 words)
- Subheadline (2 lines)
- 3 benefits (outcome-focused)
- Primary + secondary CTA
- 5 FAQ
Tone: [direct / playful / aspirational]3. Launch posts
Write 5 tweet variants for [name] launch:
- Tweet 1: hook story (why I built)
- Tweet 2: feature demo (1 line + screenshot)
- Tweet 3: pricing + free tier
- Tweet 4: ask for feedback
- Tweet 5: numbers (signups, day-1 MRR)
Tone: Pieter Levels-style — direct, no fluff4. Pricing test
Analyze 3 pricing models for [product]:
A: $9/mo unlimited
B: $0 + $0.10/generation
C: $29/mo + 100 credits/mo + $0.05/extra
For each:
- LTV estimate
- Churn risk
- Volume needed for $10K MRR
- Global appeal5. Churn email (Claude)
Write email to user who canceled [product]:
- Acknowledge feedback
- 1 specific question: "What made you cancel?"
- Offer 50% off 3 months if retry
- Short sign-off
Tone: human, not corporate
Length: <100 words08 Common pitfalls
🚨 7 mistakes
1. Over-engineer before validate → 3 months dev, 0 users
2. Pricing too low → hard to scale, burnout
3. No audience built first → launch into void
4. Niche too broad → "AI design tool" loses to "AI LinkedIn headshots"
5. Forget billing edge cases → users abuse free tier → margin negative
6. No churn tracking → MRR fake-grows, churn eats it
7. Skip API cost optimization → margin < 50% is alarm
09 Global founder advantages
| Factor | Emerging markets | US/EU |
|---|---|---|
| Engineer cost | $800-3K/month | $8-15K/month |
| Cloud cost | Same ($50-500) | Same |
| Time to ship | Same (AI tool) | Same |
| Lifestyle cost | $500-1.5K/month | $3-8K/month |
| Margin runway | 3-5x | 1x |
→ Emerging market founder can live on $2K MRR while US founder needs $10K+.
Payment stack
| Layer | Tool | Note |
|---|---|---|
| Customer charge | Stripe Atlas | Need US LLC (Delaware) — $500 setup |
| Paddle | Merchant of record, no US company needed | |
| Lemon Squeezy | Acquired by Stripe 2025 | |
| Founder pay-out | Wise | Most popular |
| Payoneer | Best for marketplaces | |
| PayPal | Higher fees |
10 Practice exercises
✍️ 3 levels
Level 1 — 1 week
- Pick 1 niche from list of 20
- Validate: landing page + 50 signups
- Use Carrd + Tally form
Level 2 — 1 month
- Build MVP (Next.js + Supabase + Replicate)
- Launch ProductHunt
- Target: 5 paying customers
Level 3 — 6 months
- $1K MRR
- 100 paying customers
- Build-in-public ≥50 followers/week
11 Continue reading
- Chapter 1 — Solo Studio — pipeline ad
- Chapter 3 — Virtual Influencer — combine product
- Chapter 5 — Sora 2 & TikTok — viral distribution
- Chapter 7 — Toolkit
- Chapter 9 — 30-Day Roadmap — execute now
Final word
"Pieter Levels isn't a genius. He shipped 70+ failed projects before PhotoAI worked.The hardest part isn't code (AI does that).It's persisting when project 60 fails."