When to open: You want to see how AI Coding companies actually do it(not how to use them) / want to start an AI Coding startup / want to understand how Lovable achieved $50 million ARR in 6 months. This wiki organizes Shensi Circle's 39 AI Coding company case studies into three sections: "Product & Technology / Growth Playbooks / Valuation Data".
One-sentence core takeaways:Cursor "training as product" anti-framework + $50 billion valuation + Lovable 6 playbooks: 1 million users in 10 months + $50 million ARR in 6 months + Cursor/Perplexity/Lovable all chose the "anti-framework" route + The 80→100 make-or-break line.
0. AI Coding company landscape (2025-2026)
| Type | Representative | Positioning |
|---|---|---|
| In-IDE AI(Editor-first) | Cursor / Windsurf / Codex | VSCode replacement |
| No-code AI site building(Output-first) | Lovable / Bolt / v0 / Readdy.ai | Natural language → complete application |
| CLI Agent(Terminal-first) | Claude Code / Codex CLI / Devin | Terminal / Agentic |
| Collaboration platform(Cloud-first) | Replit / GitHub Codespaces | Browser as development environment |
1. Cursor —— Training as product at $50 billion valuation
See Shensi Circle / AI Agent Company Case Library §1 + extensions in this section:
1.1 Core data
| Dimension | Data |
|---|---|
| Valuation | $50 billion (2026-Q1) |
| Composer model | Nearly on par with top frontier models / 4x token efficiency |
| Training stack | Thousands of GPUs / MoE / MXFP8 microscaling / Blackwell 3.5x |
| Environment | microVM (same as production cloud agents) |
1.2 Anti-framework route
"Uncovering the technical inside stories of Cursor, Perplexity, and Lovable: Why they all chose the 'anti-framework' route"
No LangChain / AutoGen → write the underlying layer in-house → more controllable and faster.
1.3 Chief designer interview —— AI-native design
(Original document:Interview with Cursor's chief designer: How to disrupt traditional interaction paradigms with AI-native design)
1.4 Cursor CEO talk
"The next era of AI coding" —— 2026 CEO talk
1.5 Cursor multi-model routing
"When $50 billion Cursor runs ads for Kimi"(2026-03)
Implication: Cursor trained Composer in-house, but still keeps multi-model routing open → user experience > model closed loop
2. Lovable —— $50 million ARR in 6 months
Shensi Circle / AI Coding Company Case Studies (Cursor/Lovable/Bolt/Devin/Replit) source:weixin/.../深度分析Lovable*
2.1 Core data
- Reached $50 million ARR in 6 months
- Acquired 1 million users in 10 months
- Traditional growth strategies have completely failed(direct quote from head of growth)
2.2 6 new playbooks (first public share by head of growth)
(See original document for details:Lovable head of growth first public share: Why traditional growth strategies completely failed, and these 6 new playbooks got us 1 million users in 10 months)
Anti-traditional growth:
- No paid traffic buying (traditional paid ads)
- No SEO keyword bidding
- Heavy UGC + showcase + community
2.3 Lovable + its Indian variant
"After Lovable, top Silicon Valley VCs invested $15 million in an Indian Vibe coding product"
3. Bolt / v0 / Readdy.ai —— Output-type Vibe Coding
3.1 Readdy.ai (Chinese team / Lanhu + MasterGo)
Shensi Circle / AI Coding Company Case Studies (Cursor/Lovable/Bolt/Devin/Replit) source:weixin/2025-06/华人团队Readdy.ai
Key data:
- Nearly $5 million ARR 4 months after launch
- Behind it: Lanhu + MasterGo team
- Multi-page website generation(not single-page)
- Code is production-ready(HTML / React / Vue)
Paradigm judgment:
"Readdy.ai showed me that real change is happening — the entire digital product development paradigm may face a fundamental shift. "
"Turning design from an 'operational skill' into an 'expressive ability'. Anyone who can clearly articulate their ideas can quickly get professional-level design results. "
3.2 Bolt / v0
- a16z quietly invested in a WYSIWYG APP development company(2026)
- v0.dev(Vercel) → Lao Huang's "Go Fundamental" (keynote on behalf of v0.dev) route
4. Cognition Devin / Claude Code / CLI Agent
4.1 Devin triangle
"After Cursor, Devin, and Claude Code, another dark horse in AI coding is rising fast"
3 major coding Agent paradigms:
- Cursor: IDE augmentation
- Devin: Autonomous Agent (Run-and-forget)
- Claude Code: CLI collaboration
4.2 Latest prediction from the father of Claude Code
"After programming is solved, what's next?"
See Claude Code Best Practices (Cal Rueb · Anthropic Internal Talk) + Claude Code Workflows (cross-author topic synthesis)
4.3 Codex series
- Cursor + Codex solves it all at once(December milestone / author of Cilegeci)
- Former GitHub co-founder raised $17 million from a16z: to build Git for the Agent era
5. Replit / Cloud Coding
5.1 Replit deep analysis
"Deep analysis | The future of AI Coding and the rise of Replit"
5.2 Replit case: Zero engineers to build a Customer Success VP
"SaaStr built an AI Customer Success VP with Replit, zero engineers, cut headcount by 70%"
Implication: Replit is not just an IDE, it is "an Agent operating environment". SaaS companies build Agents in-house with Replit → replacing human VPs.
6. Open-source AI coding tools
- An open-source AI coding tool raised $32 million, what makes it different from Cursor?(2025-08)
- The two tracks YC most wants to invest in this year were both tackled by this open-source project
- Path:Open source + protocol layer → anti-Cursor closed source
7. Rise of Vibe Coding / YC promotion
7.1 YC partner's 15 secrets for Vibe Coding
"YC partner deep practice: 15 secrets to master Vibe Coding"
See Vibe Coding and Independent Development (cross-author topic synthesis):
7.2 Karpathy etymology extension
See Karpathy System (cross-author topic synthesis) §7.
7.3 10x growth in 6 months to become a Software 3.0 pioneer
"10x growth in 6 months, uncovering how this company step by step became a pioneer in the Software 3.0 era"
8. The 80→100 make-or-break line (borrowed from the author of Cilegeci)
See Aesthetics / Democratization of Creation (cross-author topic synthesis) §5.
Key insight:
"Web Coding solved products from 0 to 80, but 80 to 100 is where the product actually reaches a usable state. "
Implications for AI Coding companies:
- Cursor / Lovable / Readdy.ai all solved 0-80
- The real winner is the one that nails 80-100 (aesthetics / code quality / multi-page / multi-model routing)
9. Cross-author intersections
| Topic | This Shensi Circle wiki | Sister wikis |
|---|---|---|
| AI Coding tool stack | Company case perspective | Evolution of AI Coding Tool Stack (cross-author topic synthesis)(synthesis)/ newtype · AI Coding Practice and Tool Stack Evolution / Wang Kai AI Coding Practice: From Cursor to Claude Code Terminal Workflow |
| Cursor training as product | Company technical inside stories | Claude Code Workflows (cross-author topic synthesis)(Cursor + Claude Code workflows) |
| Vibe Coding | YC / Lovable / Readdy | Vibe Coding and Independent Development (cross-author topic synthesis) / Karpathy System (cross-author topic synthesis) |
| Aesthetics and democratization | Readdy.ai case | Aesthetics / Democratization of Creation (cross-author topic synthesis) |
10. How to use it —— AI Coding company paths by identity
| Who you are | Primary reference | Starting point |
|---|---|---|
| Building IDE-type AI Coding | Cursor training as product + anti-framework | RL + microVM + custom kernel |
| Building No-code AI site building | Lovable / Readdy.ai | 6 playbooks / anti-traditional growth / multi-page |
| Building CLI Agent | Devin / Claude Code | Run-and-forget / Agent engineering |
| Building collaboration platform | Replit | Zero-engineer case / Agent operating environment |
| Building open-source AI Coding | $32 million open-source tool / YC two tracks | Protocol layer / open source |
| Going from 80→100 | Author of Cilegeci | Aesthetics + code quality + multimodal |
General disciplines:
- Training as product(Cursor)
- Anti-framework —— write the underlying layer in-house (Cursor/Perplexity/Lovable consensus)
- Reaching $50 million ARR in 6 months(Lovable) is the speed benchmark for new-generation AI Coding companies
- 0-80 vs 80-100 —— the real win is in the latter
- Aesthetics > model(Xiaohongshu consensus, applicable to AI Coding products)
11. Reference list
- Shensi Circle author complete works entry —— complete works entry
- Main sources:
sources/learning/深思圈/~39 pieces including "Cursor" / "Lovable" / "Bolt" / "Devin" / "Replit" / "AI Coding" etc. - Evolution of AI Coding Tool Stack (cross-author topic synthesis) —— synthesized AI Coding tool stack
- newtype · AI Coding Practice and Tool Stack Evolution —— newtype perspective (user perspective / 11 tools eliminated)
- Wang Kai AI Coding Practice: From Cursor to Claude Code Terminal Workflow —— Wang Kai hands-on practice (terminal workflow)
- Shensi Circle / AI Agent Company Case Library —— Cursor / Devin detailed Agent dimensions
- Vibe Coding and Independent Development (cross-author topic synthesis) —— Vibe Coding independent development
- Aesthetics / Democratization of Creation (cross-author topic synthesis) —— aesthetics / democratization of creation (Readdy case)
Sources and related materials
- https://mp.weixin.qq.com/s?__biz=Mzg3NDc2MjQxMg==&mid=2247491282&idx=1&sn=83dce462b2a3158fbe3c2d310235b095&chksm=ceca8711f9bd0e07208dbb1b6b925268005426b516beaf2fc666b7de7e8610fdec32044bb810
- https://mp.weixin.qq.com/s?__biz=Mzg3NDc2MjQxMg==&mid=2247488259&idx=1&sn=dff0029da13d2c8d3f1ab6d9c9b9772e&chksm=ceca8ac0f9bd03d6db571de4e4d77aba2dc2f232657824f7e552af2ca60da0650d1e4f2f59b9
- Shensi Circle author complete works entry
- Shensi Circle / AI Agent Company Case Library
- Shensi Circle / AI Primary Market VC Perspective (a16z + Sequoia + YC)
- Shensi Circle / AI Startup 0-1 Cases + Growth Playbooks
- Evolution of AI Coding Tool Stack (cross-author topic synthesis)
- Vibe Coding and Independent Development (Cross-Author Thematic Synthesis)
- newtype · AI Coding Practice and Tool Stack Evolution
- Wang Kai AI Coding Hands-On Practice: From Cursor to Claude Code Terminal Workflow