Three years of accumulation. Methodology extracted from failure lessons of personal projects including FastFacts, AiPathly, Quanta Quest, Squirrel Quick Read, and KeepRule. Rejects the lingering internet-era toxic mindset that "funding = success," and advocates "running a real business and earning cash flow."
1. Startup Mindset (memorize these 6 lines and you are set)
| # | Mindset | One-liner |
|---|---|---|
| 1 | Making money is more important than raising funding | In the internet era, very few founding teams actually made real money; by contrast, traditional business owners can earn millions in annual profit while still having time to travel and play Texas Hold'em |
| 2 | Calculate three things clearly | Customer acquisition cost / monetization efficiency / long-term operation and maintenance cost — do not launch if any one is missing |
| 3 | Do not challenge giants on their main battlefields | Large model upgrades will not eat up any startup opportunities; only combinations of hardware + systems + ecosystems integrated with AI will eat opportunities |
| 4 | Find people who are willing to pay | Do not look for users; look for people willing to pay more than your customer acquisition cost |
| 5 | Run multiple projects in parallel | The domestic market is crowded on low-ROI opportunities, and the AI dividend window is short; finish a project, make money, maintain it at low cost, then move on to the next one |
| 6 | Going global requires localization | Launch is 10 times slower, but the model is far better after PMF; the domestic market gets more competitive by the day |
2. Core stance on "funding vs. running a business"
Reject lingering internet-era toxic mindsets(Typical post from 2023-04-26):
- Only one company successfully goes public out of thousands of projects; out of a base of 50,000 people, fewer than 100 actually pocket a million
- Several VC friends said, "For AI startups, Beijing is the first choice because all the talent is in Beijing" → Wang Kai rebuts: the implicit premise is building a huge business, but how many funding announcements were there even in 2023?
- "Successful funding" only gets you a seat at the card table; whether your hand is a dud has not yet been validated
- Difference between Beijing and Hangzhou: Beijing pursues scale (not innovation, just bigness), while the South pursues making money (regardless of size)
- Beware the subconscious belief that "you have to build something huge to be successful"
Corresponding business perspective:
- Capital advantage is no longer important (VCs are unemployed and have no money)
- Won't competitors squeeze you if you don't accumulate capital advantage? Impossible — nobody has money
- An indie-developed product with 30,000 RMB monthly revenue is already rare, but that only supports one engineer → you must build slightly larger projects
- Per-item cost accounting: product name communication cost × 1 million users = 100,000 RMB hidden cost, so a product name is worth spending several days refining
3. Accounting model: cost sensitivity of internet projects
3.1 One-time costs vs. long-term O&M costs
- ✅ If it can be built in 3–10 days and makes you excited, just go for it — one-time costs are not real costs
- ❌ Long-term O&M costs (cloud, tokens, IPs, registered phone numbers, etc.) are the real costs
- ❌ Sunk costs are not costs; dare to cut losses
3.2 ARPU model vs. instant-pay model
ARPU model(calculate total value over an extended period):
- Applicable to: long-retention products; you can dare to build even if 3-year total value is 500 RMB
- Focus on: retention and renewal data
- Problem: the math does not work when traffic is expensive
Instant-pay model(every user must pay back immediately):
- Single payment must be > total ad-spend cost (e.g., if ad spend is 50 RMB, payment must be > 50 RMB)
- Applicable to: high-intent needs such as paper-writing help, AI PPT tools
- Focus on: traffic arbitrage opportunities, single-transaction payback
- Wang Kai's lesson from losing 200 USD with 0 users: assuming all high-ARPU models work for instant pay
3.3 Single-engineer cost accounting
- A full-stack engineer's monthly salary is 30,000 RMB → a product that can support one engineer = 30,000 RMB monthly revenue
- Minimum monthly cost for a small team (product + growth + engineering) is 80,000 RMB → requires 13,334 monthly paying users
- Conclusion:You must build slightly larger projects; the probability of succeeding at 5 small projects in a row is lower than succeeding at one slightly larger project
4. Traffic dividend evolution timeline
graph LR
A[2023-Q1 套壳 GPT 红利] --> B[2023-Q2 SEO 站群红利]
B --> C[2023-Q3 国内大模型扶持流量]
C --> D[2024 Chrome 插件红利<br/>商业价值放宽]
D --> E[2024 SEO 衰落<br/>Google AI Overview 减点击]
E --> F[2024-2025 海外 newsletter / Reddit / YT 推荐]
F --> G[2025 Agent token 红利<br/>Devin / 深度研究]
G --> H[2026 OpenClaw / Chrome MCP<br/>百倍 token 消耗时代]
Principles for handling each wave of dividends:
- Enter fast (validate quickly)
- Do not over-build (dividend periods are traffic flywheels, not product-depth wars)
- Be ruthless when hitting the ceiling (exit before the exponential decay phase)
- Wrapper products must prepare a contingency plan for "notifying users after being banned" (SMS is not enough; use WeChat, QR-code auto-login)
5. The "four great mountains" facing AI entrepreneurs
graph TD
A[小微 AI 创业者] --> B[硬件层<br/>苹果 / OPPO / 特斯拉]
A --> C[系统层<br/>iOS / Android / Windows]
A --> D[平台/生态层<br/>Chrome / 微信 / Slack / Notion]
A --> E[大模型层<br/>OpenAI / Anthropic / Google]
A --> F[大公司应用层<br/>Salesforce / Adobe / 美图]
B --> G[做什么能避开 Siri 的射程?]
C --> H[避开 Google labs 的小应用射程]
D --> I[只做大平台不愿做的边界]
E --> J[Claude 直接出 MCP/Skills/Code 替代独立产品]
F --> K[大公司有数据/渠道/品牌全方位优势]
Practical advice:
- Do not build things on the short Siri path (any shallow-interaction app has no value)
- Do not build things that phone makers will definitely include in their systems (photo albums, notes, translation)
- Find niche points with special needs that "hardware + systems + platforms + big companies" are all unwilling to touch
- Turning tools into services and pure software into hardware-software integration are phased breakthrough directions
6. Entry strategies for different groups of people
| Group | Wang Kai's advice |
|---|---|
| Funding-seeking entrepreneurs with internet backgrounds | Abandon funding superstition, build cash-flow businesses, run multiple projects in parallel |
| Mid-to-senior executives at big companies | Difficult to transition; the "being needed" mindset cannot produce small projects |
| Engineers / full-stack developers | Learn ad buying, learn accounting; indie development is not just about writing code |
| Product managers | Learn AI Coding (Cursor → Claude Code), break dependence on "submitting requirements" |
| Designers | Google Stitch / v0 / Lovart replace 60% of your work; either move into brand strategy or become a micro-developer |
| VCs / investors | Use AI more; don't just talk to people without getting hands-on |
| Operations / marketing | Anthropic's one-person marketing team running the entire marketing funnel is the model |
7. Several typical ways AI products die
Failure modes Wang Kai has personally experienced or observed:
- WhatsApp push bot(Wang Kai's personal postmortem): chose the wrong track; overseas push is directly covered by IMs
- AI e-commerce guide site(Project before FastFacts): wrong content generation model, token consumption math did not work
- MJ for Feishu: open source brought distribution but could not support the team
- Quanta Quest (consumer private-domain wiki AI search): Notion/Gmail data permission requests were blocked; security agency involvement in evaluation dragged out time to death
- AiPathly (AI career consulting): Doing SEO for an innovative product was completely mismatched (search volume had not even been established)
- Domestic SaaS category: low average deal size + heavy relationship/FB culture + renewal rate is meaningless
Common lessons:
- Accounting done wrong (most cases)
- Die within 3 months after launch (indie devs should not give a project 1 year of patience)
- Ostrich mentality (believing building a good product is enough, not thinking about traffic / payment)
8. SEO/SEM offense and defense evolution
2023 dividend period
- AI-generated site networks → 0.6 RMB per article
- Google long-tail keywords lacked information → mass AI production seized share
- Sanlian Lifeweek case: content library was too literary; Claude auto-generated TDK + tags, and index volume doubled in one month
2024 turning point
- Google SGE (Search Generative Experience) replaced featured snippets
- Baidu full AI search (10% of traffic → will inevitably rise)
- Rise of the AEO concept (getting AI search engines to cite your brand terms)
- glarity.app model: every user uses the plugin to summarize once → automatically generates SEO landing pages, 3 million monthly traffic
2025+ decline
- Referral traffic from AI search citing websites dropped straight by 34.5% (Ahrefs data)
- Google CEO Pichai: search will definitely change significantly
- New strategy: fight to get your own "core search terms (brand terms)" to appear in AI summaries
9. Overseas vs. domestic markets
| Dimension | Domestic | Overseas |
|---|---|---|
| Launch speed | 10 times faster | Slow, requires rebuilding trust |
| Model after PMF | Crowded, low ARPU, hard to renew | Model is far better, high subscription rate |
| Willingness to pay | Continuously declining | Consistently high |
| Big company range | Small companies have no way to survive | Niches big companies are unwilling to touch have room |
| AI tools | Inevitably free (DeepSeek, Qwen) | Can still charge for a while |
| B2B / SaaS | Heavy relationship/FB culture, hard to do | US SaaS has single rules, AI substitution is clear |
Conclusion: All projects should be built for overseas, the domestic market gets more competitive by the day.
10. AI application layer "three-step approach" (Wang Kai's long-term framework)
graph LR
A[Step 1: 用 AI 红利切入<br/>低门槛 + 快验证] --> B[Step 2: 获取流量 / 规模优势<br/>趁红利期攒护城河]
B --> C[Step 3: 延伸业务壁垒<br/>让 AI 占比降到 30% 以下<br/>构建非 AI 护城河]
The opposite: simply connecting an API + picking a niche scenario to solve one specific problem = no moat, eaten by the next version.
11. Disciplines for indie developers
- Building a product is like making a chair: when there are no chairs in the world yet, don't worry about comfort when making the first one
- Big companies give 1 year of patience; indie developers live or die in 3 months
- After validating that "people want the chair," then add armrests
- Projects will most likely die; 90% don't make much money
- Don't have an ostrich mentality: burying your head in building without thinking about traffic, payment, expansion
- Launch at least 2–3 projects at the same time; finish one, make money, maintain it at low cost, then look at the next
12. List of products Wang Kai has personally worked on
| Project | Time | Result | Key lesson |
|---|---|---|---|
| enjoy food e-commerce | Before 2018 | Company experience | Raised 100 million USD but the founding team did not make money |
| Rishiqing | Before 2018 | Company experience | Painful internet SaaS transformation |
| AI e-commerce guide site | 2023 | Shut down | Early token cost math did not work |
| FastFacts(SEC announcement crawler) | 2023 | Shut down | 300,000 RMB ceiling on carried capacity; quant teams were 5 seconds faster than us |
| MJ for Feishu | 2023 | Open source | Good traffic but no money |
| AiPathly(AI job report) | 2024 | Shut down | SEO mismatch for an innovative product |
| Quanta Quest(Consumer wiki AI search) | 2024 | Shut down | Blocked by data permissions |
| Squirrel Quick Read(PC WeChat Official Account reading + AI information sources) | 2024 | Continuously operated, supports the founder | Domestic consumer scale is limited but retention is good |
| AI smoking cessation product | 2025-2026 | Continuously operated | "Treat AI as a coach" |
| RSI AI investment strategy | 2025-2026 | 2–4% monthly returns | See Wang Kai RSI AI Investment Strategy and the "Good Asset, Good Price, Good Volatility" Framework |
| KeepRule(Investment principles wiki) | 2026 | Demo launched | AI Coding self-running database |
13. Quotes from Wang Kai
"The internet industry is a game for gamblers; we employees got lucky and got to play with high salaries for a while."
"Successful funding only gets you a seat at the card table; whether your hand is a dud has not yet been validated."
"Large model upgrades won't eat up any startup opportunities; only hardware + systems + ecosystems combined with AI will eat them."
"We should consume more tokens."
"Finish a project, make money, maintain it at low cost, then do the next one; don't grind on low-ROI opportunities."
"Going global is definitely a high ground that must be captured; no matter how many years of experience you have, you go back to zero, and competition is about judgment and diligence."
References
- Wang Kai RSI AI Investment Strategy and the "Good Asset, Good Price, Good Volatility" Framework - Investment strategy (the opposite of startups, monetizing cognition in the AI era)
- Wang Kai's AI Industry Judgments and Big-Company Competitive Landscape Observations - Industry judgments (large models / ecosystems / hardware competitive landscape)
- Wang Kai AI Coding Practice: From Cursor to Claude Code Terminal Workflow - AI Coding (foundational skill for entrepreneurs)
- Wang Kai's Overseas Promotion SOP and Multi-Account Matrix Field Guide - Overseas promotion SOP (hands-on overseas promotion)
- newtype AI Business, Industry, and Investment Observations - Contrast with Lao Huang newtype's small-business perspective