Basics
| Item | Details |
|---|---|
| Founder | idoubi (former Tencent engineer, quit 2023-10, based in Guangzhou) |
| Products | ShipAny (SaaS template framework), ThinkAny (AI search, 640K monthly visits), MCP.so (MCP marketplace, 1M monthly pageviews, cited by a16z), HeyBeauty (virtual try-on), and 11–20 more |
| Team size | 1 person |
| Revenue | ShipAny $10K in its first week ($10K in 4 hours on Christmas Day); other products vary; total income below his Tencent salary but “happiness multiplied N times” |
| AI stack | GPT-4o → Claude 3.5 Sonnet → Cursor (continuous tool stack evolution) |
| Source | BAAI Hub / InfoQ |
Storyline
Background: idoubi is a Chinese full-stack developer who went full-time on AI indie development in 2024–2025. Unlike most founders who go “all in on one product,” he picked the extreme rapid-experiment path — shipping 11 AI products in a year, spanning SaaS frameworks, AI search, virtual try-on, AI podcasts, and more.
Turning point: idoubi’s tool stack went through a clean three-stage evolution: GPT-4o for early prototypes, then Claude 3.5 Sonnet for better code quality, finally a full migration to Cursor (with Claude built in) to achieve “everything inside the editor.” The real turning point was ShipAny — he abstracted his repeatedly-rebuilt foundation into a SaaS template framework so other indie hackers could ship faster. The tool itself became the product. Even more remarkable: MCP.so, an MCP Server marketplace, hit 1M monthly pageviews and was cited by a16z.
Outcome: Of the 11 products, only ShipAny broke out (first-week $10K). MCP.so is doing great on traffic but monetization is still early. This perfectly illustrates the solo pattern — rapid experimentation is volume hedging against low single-product success rates. idoubi’s own summary: “you can’t earn money beyond your cognition.”
Workflow Breakdown
Minimum loop analysis
| Step | What he did | AI involvement | Notes |
|---|---|---|---|
| 1. Think it through | Picked topics from his own pain points and trending themes; spent no more than 1–2 days on each idea | Low | Direction relies on human market instinct and information asymmetry |
| 2. Write a spec | Quickly drafted MVP feature lists, no detailed design docs | Medium | Shipping 11 products a year doesn’t allow careful specs — closer to Vibe Coding |
| 3. Implement | Cursor + Claude generated code fast; ShipAny framework reused boilerplate | High (~80%) | Tool stack evolution directly accelerated implementation |
| 4. Validate | Launched on Product Hunt + promoted on Twitter/X; watched feedback and willingness to pay | Low | If someone pays, continue; if not, move on |
| 5. Iterate | Iterate on breakout products; abandon the ones that don’t break out | Medium | Iteration decisions driven by market data |
Tool stack (evolution path)
- Stage 1 — GPT-4o: early conversational coding, fast but code quality was inconsistent.
- Stage 2 — Claude 3.5 Sonnet: switching to Claude significantly improved code quality, especially with long-context understanding.
- Stage 3 — Cursor (with Claude built in): final migration to the Cursor IDE — “everything in one editor,” a qualitative jump in efficiency.
Roles one person carried
| Role | AI replacement | Notes |
|---|---|---|
| Product manager | 10% | Topic selection and direction are fully human |
| Development | 80% | Cursor + Claude generated most code |
| Design | 50% | AI for basic UI, human polish |
| Marketing | 30% | Product Hunt launches + Twitter/X, AI-assisted copy |
| Customer support | 20% | Mostly community interaction |
3 Reproducible Key Moves
- Abstract repeated work into a framework product (the ShipAny pattern): When you find yourself rebuilding the same foundation (auth, billing, deployment) on every new project, abstract it into a sellable framework. Use Cursor to organize your boilerplate, add docs and templates, and price it at $49–$199. “Selling shovels to gold miners” tends to be more stable than mining yourself.
- Cap each product’s experiment cost at 1–2 weeks: Use a “lightning loop” (1–4 hours) for the MVP and launch on Product Hunt to test the market. Stop investing if there are no paying users in the first week.
- Follow tool stack evolution and keep upgrading efficiency: The GPT-4o → Claude → Cursor migration was a real efficiency gain. Reassess your stack every quarter; migrate decisively when a new tool can cut dev time by 50%+.
Limits and risks
- Very low success rate: Only 1 of 11 products (ShipAny) broke out — about 9%.
- Lack of depth: The cost of rapid experimentation is that no individual product is deep enough.
- Domestic monetization ceiling: Indie hackers in China face higher paying-user acquisition costs.
- Revenue still below previous job: idoubi himself admits total income is below his Tencent salary — the solo monetization ceiling in China is meaningfully lower than overseas.