When to open: You want to go global / do overseas promotion / build a multi-account matrix / have stepped into pitfalls on overseas infrastructure. This wiki integrates 3 schools of perspective (led by Wang Kai / Lao Huang’s independent products / Lao D’s domestic adaptation).
One-sentence core:Overseas launch is 10 times slower, but after PMF the subscription rate is 3–5 times higher + single-user ARPU is 2–3 times higher(Wang Kai);Nurturing accounts on overseas platforms = 8–12 weeks of credit accumulation, there are no shortcuts;Multi-account matrixization = pushing the cost per piece close to 0 → trade production capacity for probability.
0. The 3 schools of "going global"
| Perspective | Source | Model | One-sentence stance |
|---|---|---|---|
| Matrix-based go-global school | Wang Kai’s Overseas Promotion SOP and Overseas Multi-Account Matrix Practice / Wang Kai’s 6-Platform Overseas Operation SOP / Wang Kai’s Multi-Agent Browser Matrix Architecture (v1.1 Practical Guide) | 6-platform SOP + multi-browser matrix + 4-layer isolation | "Domestic is too competitive; overseas is the high ground we must take" |
| Independent product go-global school | newtype · Vibe Coding Independent Product Practice / newtype Observations on AI Business, Industry and Investment | newtype.pro + Substack + North American user base | "Pro Users definitely use multiple tools + 3 things for 2026 (bank card + US stocks + full AI access)" |
| Domestic adaptation school | dontbesilent AI Self-Media Course / 13-lesson Systematic Methodology | rap.dev + Xiaohongshu MCP + use in mainland China without scientific internet access | "All SOPs work even without scientific internet access" |
1. Cross-author consensus (5 points)
1.1 Recommendation traffic > SEO traffic (cold start for innovative products)
Wang Kai Wang Kai’s Overseas Promotion SOP and Overseas Multi-Account Matrix Practice §1 + §7:For new products, spend 90% of effort on recommendation traffic, do not do SEO.
| Type | Traffic strategy |
|---|---|
| Innovative product(e.g., AiPathly) | Must do recommendation traffic first to educate users |
| Known category(e.g., e-commerce tools) | Direct SEO competition + paid traffic |
Why AI programming / AI job-search reports cannot directly do SEO?
- Search is a lagging indicator
- Users do not know about the thing called "AI job-search report"
- Must first educate the market through TikTok / KOL promotion
- SEO is a defensive strategy to prevent others from harvesting your audience
1.2 Universal iron rules for AI writing (Wang Kai’s exclusive 7-point self-check)
Wang Kai’s 6-Platform Overseas Operation SOP §2:
Core rule:"Claude only does translation, not polishing. Require Claude not to improvise, translate colloquially according to your meaning, the shorter the better; the more formal, the deader it sounds."
Translation prompt template:
你是我的 Reddit 翻译助手。规则:
1. 我给你中文,你直接翻译成英文,不要发挥、不要润色、不要补内容
2. 用 Reddit 用户日常口语,避免任何正式书面语
3. 缩写优先(gonna, wanna, kinda, tbh, imo, ymmv, ngl)
4. 句子要短,平均 < 12 个英文单词
5. 不要 "In conclusion" "Furthermore" 这类衔接词
6. 不要 emdash(—),用普通逗号或句号
7. 第一人称强烈,多用 I, my, me
8. 偶尔可以打错字、漏标点,更像真人
7-point self-check list:
- Average sentence length < 12 words?
- No em dashes?
- No "In conclusion / Furthermore"?
- No perfect grammar?
- First person used > 3 times?
- Paragraphs < 4 lines?
- Reads like a Slack message, not an academic paper?
1.3 Universal 4-piece infrastructure set (Wang Kai exclusive)
1. 海外手机号(英国卡 45 元 + 10 欧元激活,2026-05 推荐)
2. 海外银行卡(Wise / Payoneer / PayPal)
3. 多套账号(Google / Apple / Claude / Stripe)
4. 住宅代理 + 浏览器指纹(矩阵必备)
Lao Huang corroborates(3 things for 2026):Overseas bank card + US stocks + full AI access(see newtype Observations on AI Business, Industry and Investment §8).
1.4 Matrixization is the new paradigm (consensus + 50x scale difference)
| School | Matrix scale |
|---|---|
| Wang Kai | 50 channels / 500 Shorts per day(10 computers × 5 browsers × 1 Google account per 1 channel) |
| Lao Huang | Intensive cultivation of 4–5 platforms(newtype.pro + YouTube + Substack + X) |
| Lao D | 1–2 core platforms(Douyin + Xiaohongshu) |
1.5 Agents autonomously register accounts (workable in 2026)
Wang Kai Wang Kai’s Overseas Promotion SOP and Overseas Multi-Account Matrix Practice §3:
- AgentMail (YC S25) agent-dedicated email
- Chrome MCP controls a real browser
- reCAPTCHA passed (just click the checkbox to pass)
- HN new account restrictions (Show HN posts rejected, can only comment to build karma)
Conclusion:Platforms treat everyone the same; you must follow the rules to nurture accounts.
2. Explicit divergences (3 points — not smoothed over)
2.1 Divergence 1: Go global or stay domestic?
| Perspective | Stance |
|---|---|
| Wang Kai(Strong go-global advocate) | "Domestic is too competitive; overseas is the high ground we must take" —— After PMF, subscription rate 3–5x / ARPU 2–3x / but startup cost 10x |
| Lao Huang(Overseas independent products) | North American user base —— Stripe / tax compliance / Substack subscription economy |
| Lao D(Domestic school) | "All SOPs work even without scientific internet access" —— DeepSeek + Monica + rap.dev + Douyin 2 million views |
Real divergence:Identity determines the path(see AI Self-Media and Content Engineering (Cross-Author Thematic Synthesis) §2.1).
2.2 Divergence 2: Matrix probability bet vs intensive cultivation of account assets?
| Perspective | Stance | Quantity |
|---|---|---|
| Wang Kai(Matrixism) | Multi-agent, multi-browser, multi-account | 5+ browsers / 50 channels |
| Lao Huang(Intensive cultivation) | Cross-platform distribution but intensive cultivation on each platform | 4–5 platforms |
| Lao D(Single platform) | Intensively cultivate 1 account on a single platform | 1–2 |
Real divergence:
- Wang Kai does "production capacity × probability" (out of 50 channels, 1–2 will always produce a viral hit)
- Lao Huang does "account assets × compound interest" (1 year of accumulation = unreplicable asset)
- Lao D does "personal IP trust flywheel"
2.3 Divergence 3: Is overseas paid traffic worth it?
Wang Kai Wang Kai’s Overseas Promotion SOP and Overseas Multi-Account Matrix Practice §9.2:Pitfall lessons:
- Overseas Google Ads / Facebook Ads / TikTok Ads
- Wang Kai’s team tried 200 USD of overseas paid traffic for a Chrome extension → 0 users
- "Small and micro teams do not have the capability for overseas paid traffic"
Countermeasure: Take the "traffic validation" path:
- First create ad creatives
- Run them to a "nonexistent" product page (placeholder)
- Collect customer acquisition cost
- Decide whether to build the product
3. Wang Kai's 6-Platform SOP Quick Reference (Exclusive)
Wang Kai's 6-Platform Overseas Operation SOP:
| Platform | Cold Start | Core Strategy | Red Lines |
|---|---|---|---|
| 8 weeks | niche subreddits / 1:9 Golden Ratio / 4 Types of Tactics (Research / Creative Assets / Product Philosophy / Public Sentiment) | shadowban / Multiple accounts on the same IP / Uppercase emoji combinations | |
| HackerNews | 8-week account warming | New accounts only post comments / New accounts cannot post Show HN / Only allowed after 8 weeks / 4 high-upvote paradigms | Asking friends to vote / Frequent refreshing / Re-Showing the same product |
| Twitter / X | 90-day real-person accounts | 50% projects / 30% industry / 20% personal / Reply to 50–200 reposts that have not gone viral | Same IP means certain death / 7-day shadow review / Cross-account coordinated interaction detection |
| YouTube Shorts | 90 days per single channel | 5 themes × 5 languages × 2 styles = 50-channel matrix / The first 3 seconds decide everything | Completion rate / Matrix anti-cheat relies on fingerprints |
| ProductHunt | 12-hour anti-rush-to-leaderboard period | One-off promotion / Real user operation | Votes solicited from domestic China do not count |
| Dev.to | cross-post matrix | Long-form articles + SEO-friendly | Coordinate with Medium / Substack |
3.1 HN Show HN Template
Title: Show HN: [Product] – [one-line value prop, max 80 chars]
I'm [your role]. Built [product] because [specific personal pain].
What it does:
- [specific capability 1]
- [specific capability 2]
- [specific capability 3]
Tech stack: [key tech]. Architecture decisions worth noting:
[1-2 design tradeoffs that other devs will appreciate]
Pricing: [be transparent — free tier / paid / open source]
Honest about limitations: [admit 1-2 known gaps]
Would love feedback on [specific question].
Source code: [github link if applicable]
Demo: [link]
3.2 Twitter Recruitment Tweet Template (Tested by Wang Kai)
Hiring: [Role] for [project]
[1 sentence about what we're building]
What we need:
- [skill 1]
- [skill 2]
- [working style: remote/async/etc]
What you get:
- [comp range — be specific]
- [equity if applicable]
- [interesting tech challenge]
Reply or DM. Retweets appreciated.
Prerequisite: Effective recruitment tweet = Over 1 year of reputation on a real-person account × Concentration of target audience among followers.
3.3 YouTube 50-Channel Matrix Differentiation
5 themes × 5 languages × 2 styles = 50 channels:
| Dimension | Example |
|---|---|
| Theme | AI tool reviews / Market interpretation / Historical stories / Programming tutorials / Psychology |
| Language | English / Spanish / Portuguese / Hindi / Arabic |
| Style | Serious explanation / Funny twist / Data visualization / Real-person on-camera appearance |
Each channel is independent: channel name + logo + banner + bio + voiceover timbre + thumbnail style + posting schedule.
3.4 Hook Template (The First 3 Seconds Decide Everything)
- Contrast type:
"Everyone says [common belief]. They're wrong. Here's why." - Data type:
"This single change made [metric] go [direction] [number]." - Question type:
"Why does [common phenomenon] actually happen?" - Demo type:
"Watch what happens when [action]."
4. Wang Kai's Multi-Agent Browser Matrix Architecture (Exclusive Engineering Implementation)
Wang Kai's Multi-Agent Browser Matrix Architecture (v1.1 Practical Guide) v1.1 Complete Practical Guide:
4.1 Core Definitions
身份 (Identity) = 一个浏览器 app + 一套登录态 + 一个住宅 IP + 一个 cdp-proxy 进程
↓
工种 (Role) = 一个 Claude Code agent(独立工作目录),跟同身份其他 agent 共享浏览器实例
4.2 Port Pair Allocation (By Identity Dimension v1.1)
| Identity | Browser app | proxy port | chrome port | Residential proxy exit |
|---|---|---|---|---|
| identity-A | Chrome | 9401 | 9222 | US-East |
| identity-B | Arc | 9402 | 9223 | US-West |
| identity-C | Brave | 9403 | 9224 | UK |
| identity-D | Edge | 9404 | 9225 | DE |
| identity-E | Vivaldi | 9405 | 9226 | JP |
4.3 Four Layers of Isolation
- Browser isolation(Different apps)
- Account isolation(Independent Google / Independent user data directory)
- IP isolation(Residential proxy exit)
- Fingerprint isolation(Multilogin / GoLogin / AdsPower)
4.4 Browser Fingerprint Tool Selection
| Tool | Price | When to use |
|---|---|---|
| GoLogin | Free for 3 profiles / $24/mo | Starting out / ≤ 3 identities |
| AdsPower | Free for 5 profiles / $5.4/mo | Domestic teams going overseas |
| Multilogin | Starting at $99/mo | TikTok / X strict crackdown period / ≥ 5 identities |
| BitBrowser | (Price not provided) | Alternative option |
4.5 Residential Proxy Selection
| Provider | Price | Notes |
|---|---|---|
| Bright Data | $8.4/GB | Established provider, largest IP pool |
| Soax | $7/GB | Decent cost-performance |
| IPRoyal | $7/GB | Low starting price |
| Smartproxy | $7/GB | Beginner-friendly |
Datacenter proxies are absolutely prohibited —— X / TikTok / Reddit ban accounts immediately upon registration on datacenter IPs
4.6 Iron Rule of Geographic Consistency
- identity-A = US East Coast user → US-East residential IP → Time zone
America/New_York→ All platform accounts post on EST time - Geographic inconsistency is the easiest risk control trigger to catch
4.7 CDP Security (Must Read)
- Default listens on 127.0.0.1
- ❌ Never change to
0.0.0.0 - ❌ Never perform
docker -p 9222:9222port mapping - ❌ Matrix browsers are prohibited from logging into important personal accounts (personal Gmail / work email / bank accounts)
5. Going-Overseas Infrastructure Checklist (Wang Kai Exclusive)
Wang Kai's Overseas Promotion SOP and Overseas Multi-Account Matrix Practice §2:
graph TD
A[出海基础设施] --> B[支付层]
A --> C[账号层]
A --> D[流量层]
A --> E[审核层]
B --> B1[Stripe / Lemon Squeezy / Gumroad]
B --> B2[Wise / Payoneer / PayPal]
B --> B3[Apple / Google Play 海外账号]
C --> C1[英国卡 45 元 + 10 欧元激活]
C --> C2[AgentMail Agent 专用邮箱]
C --> C3[多套 Google / Apple / Claude 账号]
D --> D1[Reddit / Twitter / YouTube]
D --> D2[HackerNews / ProductHunt]
D --> D3[Dev.to / Medium / Substack]
E --> E1[Google OAuth 安全审核]
E --> E2[Apple Developer 99 美金/年]
E --> E3[Stripe Atlas 公司 1 千美金起]
5.1 Hidden Costs of Going Overseas
| Dimension | Pitfalls |
|---|---|
| Data security review | Connecting to Google Workspace API → Similar to a vehicle inspection / Quanta Quest got stuck here |
| Paid traffic capability threshold | $200 overseas Chrome extension paid traffic → 0 users |
| Credit system | HN / Reddit both require account warming / At least 3–6 months |
| Payment threshold | Stripe Atlas $1,000 / Apple Developer $99 / Wise transfers to ZA Bank blocked |
5.2 Overseas Promotion Pitfall Avoidance Checklist
| Pitfall to avoid | Reason |
|---|---|
| ❌ Directly posting soft ads on Reddit / HN | Uniformly rejected |
| ❌ Directly posting Chinese translations overseas | Severe cultural mismatch |
| ❌ Soliciting domestic votes for ProductHunt | Does not count, total waste |
| ❌ Overseas paid traffic without localized assets | ROI will never turn positive |
| ❌ Large-scale operation on a single account | Total loss if the platform bans the account |
| ❌ Using Google Voice / Skype digital accounts | Detected and banned |
| ❌ Using mainland China bank cards for ZA Bank remittances | High block rate |
6. Lao Huang's Independent Product Going Overseas (Exclusive)
newtype · Vibe Coding Independent Product Practice:
6.1 Individual Business Experience
- US company
- Apple Developer $99/year
- Substack 200 subscribers
- newtype.pro monthly Newsletter
- Pro User multi-tool (subscriptions to Claude / ChatGPT / Cursor / Gemini)
6.2 Three Things for 2026
newtype AI business, industry and investment observations §8 ("What to do in 2026" provided by Lao Huang):
- Overseas bank card
- US stocks
- Full AI Access(Subscribe to frontier models)
6.3 Don't compete with giants on the "input box"
"Chatbots like ChatWise, if they cannot find a narrow niche to survive, will eventually be crushed by big tech companies.This is the biggest taboo in AI entrepreneurship. "
What Lao Huang built with Prompt House is not an "input box", but "Prompt management + MCP protocol orchestration" (atomized middle layer).
7. Lao D's one-click domestic publishing (exclusive)
dontbesilent AI self-media course / 13 lessons of systematic methodology §L11:
7.1 rap.dev + Xiaohongshu MCP workflow
1. 在 RAP 输入:"帮我把这条推文发布到小红书"+ 链接
2. RAP 调用:
a. Playwright 操控无头浏览器截图(确保 3:4 比例)
b. 字数多 → 左右填白边(Python Pillow)
c. 字数少 → 拉远拍 → 裁剪
3. Claude 4.5 Sonnet 给小红书写标题(自动 ≤ 20 字 / 从原文找金句)
4. 小红书 MCP server 自动登录发布
5. 还可以定时发布("1 小时后发")
7.2 Comparison of domestic vs overseas rhythm
| Platform | Domestic characteristics | Overseas characteristics |
|---|---|---|
| Rhythm | Faster rhythm but lower content quality threshold | Credit accumulation takes 8–12 weeks, no shortcuts |
| Anti-cheating | More lenient (Lao D did not emphasize account nurturing) | Extremely strict (same IP means certain death/ban) |
| Willingness to pay | Continuously declining | Consistently high |
| Moderation | Content moderation (policy sensitivity) | Data security + privacy compliance |
8. How to choose — going-global paths by identity
| Who you are | Primary method | Starting point |
|---|---|---|
| Small and micro AI teams going global | Wang Kai matrix approach | 6-platform SOP + multi-browser matrix + 4-piece overseas infrastructure set |
| Solo non-technical indie product builder | Lao Huang indie product approach | newtype.pro + Substack + Pro User + three things for 2026 |
| Domestic ordinary users without VPN access | Lao D domestic adaptation | rap.dev + DeepSeek + Xiaohongshu MCP + Douyin |
| Want to build a 50-channel matrix | Wang Kai multi-browser matrix | 4-layer isolation + fingerprint tools + residential proxies + CDP security |
| Focus on overseas content factories | Wang Kai YouTube 50 channels | 5 themes × 5 languages × 2 styles + Hook templates |
General discipline:
- For new products, spend 90% of effort on recommendation traffic; SEO is defensive
- Nurturing accounts on overseas platforms = 8–12 weeks of credit accumulation, no shortcuts
- Claude is only used for translation, not polishing(overseas) / Preserve personal style, do not eliminate the AI feel(domestic)
- Never post the same draft on two platforms
- Small and micro teams without overseas paid traffic capability should follow the traffic validation chain
- Datacenter proxies are strictly prohibited; must use residential proxies + geographic consistency
- Never expose the CDP port on 0.0.0.0
9. Reference list
Main original author wikis:
- Wang Kai: Overseas Promotion SOP and Overseas Multi-Account Matrix in Practice —— 226 posts on overseas practice / overseas infrastructure checklist / innovative vs known categories / hidden costs / pitfall avoidance checklist
- Wang Kai: 6-Platform Overseas Operations SOP —— Full 6-platform SOP / 7 self-checks for AI writing / HN Show HN template / Twitter recruitment template / YouTube 50 channels
- Wang Kai: Multi-Agent Browser Matrix Architecture (v1.1 Practical Guide) —— v1.1 multi-browser matrix / 4-layer isolation / port pairing / fingerprint tools / residential proxies / CDP security
- Wang Kai: Entrepreneurship Methodology and Business Judgment Framework in the AI Era —— "Going global requires localization" / running multiple projects in parallel
- newtype · Vibe Coding Indie Product Practice —— Lao Huang indie products / North American user base / Pro User
- newtype AI Business, Industry and Investment Observations §8 —— Three things for 2026 (overseas bank card + US stocks + full AI access)
- dontbesilent AI self-media course / 13 lessons of systematic methodology §L11 —— rap.dev + Xiaohongshu MCP / domestic adaptation
Related topic wikis (this series):
- AI Self-Media and Content Engineering (Cross-Author Topic Synthesis) —— AI Self-Media and Content Engineering (Batch 1)
- Claude Code Workflow (Cross-Author Topic Synthesis) —— Claude Code Workflow (Batch 1)
- Entrepreneurship Methodology (From 0 to 1) (Cross-Author Topic Synthesis) —— Entrepreneurship Methodology (Batch 1)
History
- 2026-05-18: First release of the Topic-dimension wiki. Synthesized and rewritten from 7 author wikis. Three disagreements are explicitly marked: going global vs staying domestic / matrix probability betting vs intensive cultivation / whether overseas paid traffic is worth it
Sources and related materials
- Wang Kai: Overseas Promotion SOP and Overseas Multi-Account Matrix in Practice
- Wang Kai: 6-Platform Overseas Operations SOP
- Wang Kai: Multi-Agent Browser Matrix Architecture (v1.1 Practical Guide)
- Wang Kai: Entrepreneurship Methodology and Business Judgment Framework in the AI Era
- newtype · Vibe Coding Indie Product Practice
- newtype AI Business, Industry and Investment Observations
- dontbesilent AI self-media course / 13 lessons of systematic methodology
- AI Self-Media and Content Engineering (Cross-Author Topic Synthesis)
- Claude Code Workflow (Cross-Author Topic Synthesis)
- Entrepreneurship Methodology (From 0 to 1) (Cross-Author Topic Synthesis)
2026-10-18 supplement: Shifting from channel matrix to growth driven by "audience, value, and feedback"
This section synthesizes interviews from 42 Chapters across different years to supplement the earlier perspective focused mainly on channels and production capacity.The earlier overseas payment multiples, platform account-nurturing durations, account and IP judgments, and fixed channel proportions are retained as the original authors' contemporary experience, not re-verified platform rules or universal growth laws for this supplement. The revenue, retention, payment, and ROI figures below should also be understood only within their respective cases, time points, and measurement scopes; do not use self-reported numbers from one market to set budgets for another product. This article is a cross-temporal knowledge distillation and does not constitute authorization for automated posting, recruitment outreach, or ad placement.
1. At each stage, first identify one primary constraint, then choose channels and organization
Growth does not start with "how many accounts should I open"; it starts by asking: Do target users not yet know about the product, not trust the product, not experience the value, not want to pay, or not come back after trying? These questions correspond to different actions. Expanding exposure before value is delivered may only increase failed experiences; continuing to pile on features when there is stable payment but no awareness may also not change outcomes. In a 2025 interview, Xie Juntao, former head of growth at OpusClip, separated cold start, conversion, retention, and insight: in the early stage find the right people; after registered traffic exists learn the payment points; after the existing base expands close the feedback loop; invest in data infrastructure according to complexity. The point is to locate the constraint, not to require every company to build a complete growth department at the same time.Original program text
ACE Studio's HF0 experience in 2025 provides another organizational approach: use time-bound single-goal sprints, shifting weekly reports from "how many features were built" to "which hypothesis produced results". However, it already had about $100,000 in monthly revenue before entering the program, making it a team with an existing value foundation; the residency environment, pricing, channels, attribution, and return traffic all changed at the same time, so it cannot be assumed that an incubator can universally scale zero-start teams tenfold. What can be borrowed is reducing attention fragmentation within a fixed cycle, while each experiment still retains guardrails for refunds, quality, retention, and costs; long working hours and high pressure cannot be copied as necessary conditions.Original program text
Gaotu's internal practice in 2024 shows that priority also depends on scope of control and result visibility: saving tutors' time may not translate into equivalent cash returns, public-domain content customer acquisition is affected by platforms and competition, while responses, Q&A, and conversion among owned leads are easier to trace to results. Zero-lead new products cannot copy "conversion first", and existing businesses should not use AI coverage rate as a substitute for closed deals. Choosing only one main thing per stage is for the sake of being able to judge causes, not to cancel other responsibilities.Original program text
2. Channel × audience × content × trust is a joint hypothesis
Pitching the same feature to different groups of people may not be the same business. Answer.AI explained in 2024 that choosing TikTok for high school students was not because the platform was inherently cheap, but because the target audience matched content habits; agencies familiar with paid traffic operations do not necessarily know how to make creatives suitable for a specific product. The experience of Indian users not understanding the magnifying glass icon shows that interface habits can also block the funnel, and low conversion cannot be directly attributed to lack of demand. Language, education systems, devices, and life scenarios should be included in research, rather than treating translated Chinese scripts as completed localization.Original program text
In ACE Studio's cross-period interviews in 2024, initially contacting over a hundred YouTubers received few responses; later, at the NAMM show, letting musicians actually try the product and build relationships yielded promotion responses. Its professional singing voice product requires users to trust that it can fit into their workflow, and live demonstrations, industry identity, and word of mouth changed the reach conditions. The team stated that overseas new paid conversion rates were higher and the share of US dollar revenue increased,but the sample simultaneously changed audience, channel, reach method, and trust conditions, so this cannot be abstracted as "overseas users are naturally more willing to pay". The ROI = 0 or 100 mentioned in the program lacks definitions, expense scope, and attribution windows, and is retained as an impressionistic description, not used as an audited return multiple.Original program text
In its early days, Opus took Podcasters as a clear audience group and found long-term partners within their networks where KOCs learned editing. The key was that partners were already using the product, willing to co-build and demonstrate it, and economic arrangements made the relationship sustainable; it was not about asking anyone with many followers to post a one-time ad on commission. A small number of genuine partners can generate both product feedback and credible dissemination at the same time. This is not mutually exclusive with a scaled KOL funnel: the former helps determine value and language, while the latter reduces repeated collaboration costs after mechanisms are clearer. Before scaling, distinguish natural users, paid collaborators, distributors, and ordinary consumers to avoid mixing them into a single "influencer user" label.
An executable channel experiment should be written as one complete sentence: "In a certain region/language/audience group, through a certain touchpoint, demonstrate a certain real delivery, so that users believe it and complete a certain action." If you only fill in the platform name and CPM, missing audience, content, and trust, test results will be hard to explain, and certainly cannot be indiscriminately expanded to a multi-account matrix.
3. What manual recruitment buys should be understanding and effective supply, not just sample size
In an interview recorded in November 2023 and released in 2024, FlowGPT self-reported that in the first two weeks it private-messaged five to six hundred people, cumulatively several thousand people. Its recruitment on Discord, Reddit, and Twitter targeted community members willing to create and consume works, not just one-time visits. For UGC platforms, users understand the gameplay better than the team, and early manual communication is also building supply. The headcount is a guest recollection rather than a reproducible threshold; small teams should care more about who actually posts after being reached, whether works are consumed, and whether creators continue updating.Original program text
In 2025, Mercor pivoted from its broad recruitment vision to domain experts urgently needed by AI Labs, because payers had clear budgets and delivery pressure. Professional forums, offline events, proactive outreach, and referrals jointly supplemented supply; referrals accounted for an important part of sign-ups, and should not be cited as "more than half" because that ratio was prompted by the host and not precisely confirmed. More importantly, after registration there are still identity verification, resume, skills, working hours, quality, payment, and dispute handling. Measuring growth by sign-ups misses what the platform truly needs: "effective experts completing qualified work." Long-term trust comes from fulfillment, not just finding contact information.Original program text
User research must also verify recruitment sources. Answer.AI once received users claimed by a third party to be highly active, but checking UIDs found they had just registered, making the previous days of interviews and needs understanding lose basis. Sample authenticity comes before interview technique: confirm usage duration, core actions, payment status, region and language, retain screening criteria, and record non-respondents and churned users. Do not only interview active community members willing to praise the product and then generalize conclusions to everyone.
After Monica acquired ChatGPT for Google in 2023, it did not force those users to upgrade directly; instead, it used part of the initial sample to help with validation while building its own customer acquisition capabilities. The self-reported proportion in questionnaires of "would be very disappointed if the product disappeared" rose from 35% to 55%, but sample size, selection rules, and leavers were not disclosed, so the 40% empirical threshold cannot be treated as a universal PMF certification. Acquired users, things interviewees are willing to say, behavioral retention, and actual payment are different types of evidence.Original program text
4. Free retention, real payment, and sustained gross margin must be verified separately
“The team likes their own prototype → someone uses it for free and comes back → someone is willing to pay continuously” are three assumptions. Answer.AI believes commercialization must be validated early, and inference costs also make pure free user scaling harder to sustain long-term. However, different categories should not be judged by the same next-day retention or payment rate: AiPPT’s low-frequency non-professional users may only return when they have a need; ACE serves professional production workflows; NieTa’s sustained demand may be raising characters rather than creating comics. Low frequency does not automatically equal failure, and high frequency does not automatically equal payment value.Original program text Original program text
FlowGPT also reminds that for entertainment consumption, one must separate single-work/character retention, category retention, and platform retention. If a user switches to another Bot but still comes back weekly for content, it should not be directly counted as the platform completely losing value; conversely, platform activity does not prove that a particular character relationship is stable. UGC platforms also need to look at creator retention and re-creation, to prevent consumption-side growth from masking supply contraction.
Opus’s paywall experiments emphasize first identifying what value users experience: after subtitle templates were displayed more prominently, conversion unexpectedly improved, indicating that users may value this outcome more than the team expected. The paywall interception during Arcade’s public product demo was the guest’s own purchase experience, not proof that it is optimal for all users. Pricing, benefits, prompt timing and copy must be recorded together to protect users who have already purchased; for each change, look at first-time payment, subsequent usage, cancellations, refunds and feedback simultaneously. AI cost accounting must also include free trial consumption, and cannot use only the gross margin of paying users.
ACE stated in 2025 that one price increase quickly doubled daily revenue, after which it tested different prices, trials and monthly passes. This is a short-term cash lever, and does not prove that long-term renewals are unharmed. A monthly revenue of 800,000 USD simply annualized to about 9.6 million USD also does not automatically become full-year actual receipts or renewable ARR. Mercor’s 100 million USD run rate is likewise not accumulated revenue. Growth reports must clearly distinguish cash receipts, revenue, GMV, run rate, ARR and profit, and not let headlines replace definitions.
V. Growth quality also includes ecosystem spillover and user reception capacity
Reviewing Douyin from 2018 to 2020, Yu Beichuan emphasized that the early young user base, black UI tone, curated supply and avoidance of prematurely importing broad traffic jointly established its positioning; it then introduced new vertical categories, gave exploratory exposure so recommendations could learn, and then attracted new user groups. This is an explanation based on internal experience, lacking experiments that isolate each factor, but the mechanism reminds us: acquiring new users is not just obtaining an ID; there must also be content, entry motivation and learnable feedback to receive them.Original program text
Distribution containers simultaneously change viewing efficiency and author returns. Single-column and double-column layouts cannot be compared only by single-session watch time; one must also look at ordinary authors’ first feedback, re-creation, supply diversity and user autonomous choice. Authors serve consumers both inside and outside experiments, so ordinary user-level A/B tests will have network spillovers; short-term consumption advantages do not necessarily represent a better long-term ecosystem. It is necessary to define clear segmentation, long observation windows and potential contamination, rather than exempting tests from scrutiny because they are hard to measure.
The same history also contains proxy metric distortion: OKRs for adding friends and interactions can be inflated by red envelopes, without natural relationship scenarios; changing targets near Spring Festival campaigns may neither build social connections nor perform well in user acquisition. These examples remind us that follows, shares, messages and generation consumption are all just processes, and we must return to why users come back. Complaints about insufficient stamina on NieTa may indicate stronger demand, but real purchases, repurchases and supply impacts still need to be verified; complaints or consumption cannot be directly treated as commercialization success.
Content dissemination can also harm reception: low-cost traffic causes registrations to surge, but mismatched users, overwhelmed customer service, and authors not receiving feedback will make short-term CAC look good while subsequent value deteriorates. Growth quality should include the proportion of wrong audiences, task failures, complaints, refunds, aversion and supply loss; one cannot use “there is exposure” alone to justify investment.
VI. Connecting experimental results back to product and operations completes the growth closed loop
The growth closed loop has at least five steps: propose the current constraint → write down mechanisms and metrics → run limited-budget experiments → observe results and guardrails → decide to retain, revise or stop. The one-week fast decision-making described by Mercor is premised on first clarifying what should be seen in the coming week and what to do if it is not seen; speed does not mean skipping expectations, responsibilities and failure plans. Complex or irreversible actions require higher thresholds and cannot copy the pace of reversible experiments.
Opus connects customer service issues to product priorities, fixes, and customer service follow-ups: community wishes, paid-user malfunctions, and roadmap items can be collected through separate channels, but ultimately someone must be responsible for closing the loop. Merely auto-classifying or sending an email does not mean the problem is fixed. For renewals, active cancellations must also be separated from Past Due payment failures; the 20%–30% to 40%–50% retry success rates cited are undocumented experiential figures without a stated methodology, not a growth budget. In practice, one should look at retry-eligible targets, recovered amounts, authorization status, duplicate charges, and complaints, and respect users’ already stated choice to cancel.
The Gaotu program provides an important caliber correction: the official chapter states the order rate is three to four times higher, but the main text specifically discusses a three- to four-fold increase in “per-lead efficiency” measured by GMV per lead, and separately states that the conversion rate increased by about one time, and that this was only a small-scale experiment. Operational readback should prioritize specific definitions, retaining sources, experiment proportions, baselines and windows; one cannot expand a chapter title into a company-wide revenue doubling. Covering more leads and replying to more messages are still just process metrics.
VII. Practices for migrating to product teams: first create a growth experiment card that can be read back (editorial inference)
For AI companion/character and creation products, one can first fix one region and language, one target user group, one core experience, and one channel, and not change price, entry points, content and recommendations in the same week. The card should record at least: target audience and its authenticity, existing alternatives, first value action, channel and creative versions, cost inclusions, experiment denominator and time window, follow-up visits and payments, refunds and quality guardrails, author feedback, and stop/scale-up conditions. Qualitative verbatim quotes must be traced back to corresponding samples, behavioral results must be traced back to events and payment flows, and questionnaires, demos, backend saves, launches and real user experiences should not be conflated into one layer.
Creator growth first looks at “recruited → completed work → first valid consumption feedback → continued creation”; consumer growth looks at “saw card → started interaction → received expected experience → came back again”; commercialization then looks at real purchases, renewals and gross margin. Report separately at the character level, category level and platform level; damage to relationship experience cannot be offset by short-term payment. After a channel yields good results, gradually increase audience and scale, checking each time whether the original trust and supply mechanisms still hold. A matrix is an amplifier of validated mechanisms, not a substitute for the mechanisms themselves.
English version
2026-10-18 Addition: Growth Driven by Audiences, Value, and Feedback, Beyond a Channel Matrix
This section synthesizes 42 Zhangjing interviews from different years, adding to the preceding discussion's emphasis on channels and production capacity. The earlier claims about overseas payment multiples, time spent establishing platform accounts, account and IP judgments, and fixed channel shares remain the original author's experience at the time. They are not platform rules reverified for this update or universal laws of growth. The revenue, retention, payment, and ROI figures below must likewise be understood within their respective cases, dates, and definitions. One market's self-reported figures should not become another product's budget assumptions. This is a synthesis across episodes; it does not authorize automated posting, recruitment outreach, or advertising expenditure.
1. Identify One Main Constraint at Each Stage, Then Choose Channels and Organization
Growth starts with a question, rather than with how many accounts to open: Do target users not know the product, not trust it, fail to experience its value, decline to pay, or stop returning after trying it? Each problem calls for a different action. Expanding exposure before delivering value may simply create more failed experiences. Adding features to a product that already has stable paying users but little awareness may also leave the outcome unchanged. In a 2025 interview, former OpusClip growth lead Xie Juntao separated cold start, conversion, retention, and insight: find the right people early; learn what drives payment once registration traffic exists; build feedback loops as the installed base grows; and invest in data infrastructure according to complexity. The point is to locate the constraint, not to require every company to build a complete growth department at once.Original episode
ACE Studio's 2025 experience at HF0 offers another organizational approach: use a time-limited sprint around a single goal, shifting weekly reports from “how many features we built” to “which assumption produced a result.” But the team already had approximately US$100,000 in monthly revenue before entering the program, so it had an existing foundation of value. Its working environment, pricing, channels, attribution, and returning traffic changed together. This does not show that an accelerator can generally multiply a team starting from zero by ten. The transferable practice is reducing divided attention over a fixed period, while retaining refund, quality, retention, and cost guardrails for every experiment. Long hours and high pressure should not be copied as necessary conditions. Original episode
Gaotu's internal work in 2024 shows that priorities also depend on what a team can control and whether outcomes are observable. Saving tutors' time does not necessarily turn into equivalent cash savings; acquiring customers through public-platform content is affected by platforms and competition; and responses, answers, and conversion within an existing lead pool are easier to follow through to outcomes. A new product with no leads cannot simply copy a “conversion first” strategy, while an established business should not substitute AI coverage for completed sales. Choosing one main priority at a stage helps identify causes; it does not remove other responsibilities. Original episode
2. Channel × Audience × Content × Trust Is a Joint Hypothesis
The same feature offered to different people may amount to a different business. Answer.AI explained in 2024 that it chose TikTok for high school students because the audience and content habits matched, not because the platform was inherently cheap. An agency's familiarity with paid acquisition does not necessarily mean it can produce creative suited to a particular product. Its experience with users in India who did not understand a magnifying-glass icon also shows how interface conventions can obstruct the funnel. Low conversion cannot immediately be attributed to a lack of demand. Research should include language, education systems, devices, and everyday contexts, rather than treating translated Chinese copy as completed localization. Original episode
In a 2024 interview included in this synthesis across episodes, ACE Studio described initially contacting more than a hundred YouTubers with little response. Responses to promotion followed only after it let musicians genuinely try the product and built relationships at the NAMM trade show. Its professional singing-voice product required users to believe it could fit into their workflow; live demonstrations, industry standing, and word of mouth changed the conditions for reaching them. The team reported a higher rate of new users becoming paying users overseas and an increased share of revenue in US dollars. But the sample changed in audience, channel, contact method, and trust conditions simultaneously. It cannot therefore be generalized into “overseas users are inherently more willing to pay.” The episode's ROI figures of 0 or 100 have no stated definition, cost scope, or attribution window. They are retained as impressions, not as audited return multiples. Original episode
Early on, Opus identified podcasters as a specific audience and found long-term partners in the networks where those podcasters learned editing from key opinion consumers (KOCs). What mattered was that the partners already used the product and wanted to help build and demonstrate it; financial arrangements made those relationships sustainable. This was not simply a one-off commission-based advertisement from anyone with a large following. A few genuine partners can produce both product feedback and credible distribution. This approach and a scaled key opinion leader (KOL) funnel are compatible: the former helps establish value and language, while the latter reduces the cost of repeated collaboration once the mechanism is clearer. Before scaling, distinguish organic users, paid partners, distributors, and ordinary consumers instead of combining them into one “influencer user” label.
An actionable channel experiment should form a complete sentence: “For a particular region, language, and audience, use a particular touchpoint to demonstrate a real deliverable, so users trust it and complete a particular action.” A test specified only by platform name and CPM lacks audience, content, and trust. Its results will be difficult to interpret, and cannot justify indiscriminate expansion across an account matrix.
3. Manual Recruitment Should Buy Understanding and Effective Supply, Not Just Sample Size
In an interview recorded in November 2023 and released in 2024, FlowGPT reported privately messaging five or six hundred people during its first two weeks, and several thousand in total. Its recruitment on Discord, Reddit, and Twitter targeted community members willing to create and consume works, rather than merely purchasing a visit. On a user-generated content platform, users themselves may understand the possibilities better than the team; early manual conversations also build supply. These counts are the guest's retrospective account, not a reproducible threshold. A small team should care more about who actually publishes after being contacted, whether anyone consumes those works, and whether creators keep updating them. Original episode
In 2025, Mercor moved from a broad recruitment vision toward domain experts urgently needed by AI labs, whose buyers had clear budgets and delivery pressure. Professional forums, offline events, proactive recruitment, and referrals together supplied experts. Referrals were an important part of registrations, but should not be quoted as “more than half”: the host raised that proportion in a question and it was not precisely confirmed. More importantly, registration was followed by identity, résumé, skill, hours, quality, payment, and dispute management. Measuring growth by sign-ups misses what the platform actually needs: effective experts completing qualified work. Long-term trust comes from fulfilling commitments, not merely finding contact details. Original episode
User research must also verify recruitment sources. Answer.AI once received participants whom a third party described as highly active users, only to discover from their UIDs that they had just registered. The interviews and understanding of demand developed over the preceding days consequently lost their foundation. Sample authenticity comes before interview technique: confirm usage history, core actions, payment status, region, and language; preserve screening conditions; and record nonrespondents and churned users. Interviews only with active community members who are willing to praise the product cannot represent everyone.
After acquiring ChatGPT for Google in 2023, Monica did not force its users to upgrade directly. It used part of the initial sample to help validate the product while building its own acquisition capabilities. It reported that the share of survey respondents who would be “very disappointed” if the product disappeared rose from 35% to 55%. But sample size, selection rules, and people who had left were not disclosed; a 40% rule of thumb cannot be treated as a universal PMF certification. Acquired users, what people are willing to say in interviews, behavioral retention, and actual payment are different forms of evidence. Original episode
4. Retention Among Free Users, Actual Payment, and Sustained Gross Profit Need Separate Validation
“The team likes its own prototype → people use it for free and return → people are willing to keep paying” contains three hypotheses. Answer.AI argued that monetization must be tested early, while inference costs make purely free scaling harder to sustain. But different categories should not share one next-day retention or paid-conversion target: AiPPT's infrequent, nonprofessional users may return only when a need arises; ACE serves professional production workflows; and NieTa's recurring demand may be nurturing characters rather than creating comics. Infrequent use does not automatically mean failure, and frequent use does not automatically demonstrate value worth paying for. Original episode Original episode
FlowGPT also highlights the need to separate retention for an individual work or character, category retention, and platform retention in entertainment consumption. A user who switches to another bot but still returns weekly to find content should not be counted as a complete loss of platform value. Conversely, platform activity does not prove that a particular character relationship is stable. UGC platforms must also track creator retention and continued creation, so growth on the consumption side does not conceal shrinking supply.
Opus's experiments with payment prompts emphasized first identifying the value users actually experience. Making subtitle templates more visible unexpectedly improved conversion, suggesting users may have valued that result more than the team anticipated. The paywall encountered during Arcade's public product demonstration was the guest's own purchasing experience, not proof of the best approach for all users. Pricing, entitlements, prompt timing, and copy should be recorded together, while protecting existing purchasers. Each change should be evaluated against first payments, subsequent use, cancellations, refunds, and feedback. AI cost accounting must include consumption during free trials, not just the gross margin of paying users.
In 2025, ACE said one price increase quickly doubled daily revenue, followed by tests of different prices, trials, and monthly plans. That is a short-term cash lever, not proof that long-term renewals were unharmed. Simply annualizing US$800,000 in monthly revenue gives approximately US$9.6 million; it does not automatically become cash collected over a full year or renewable ARR. Mercor's US$100 million run rate likewise does not mean that amount had already been earned in cumulative revenue. Growth reports must clearly distinguish cash collected, revenue, GMV, run rate, ARR, and profit, rather than allowing a headline to substitute for definitions.
5. Growth Quality Includes Ecosystem Spillovers and the Capacity to Serve Acquired Users
Reflecting on Douyin from 2018 to 2020, Yu Beichuan emphasized how its early young audience, black UI aesthetic, curated supply, and restraint in importing broad, untargeted traffic established its positioning. It subsequently introduced new verticals and allocated exploratory exposure so recommendations could learn, then attracted new audiences. This is an explanation based on internal experience, without experiments isolating each factor. But the mechanism reminds us that acquisition means more than obtaining an ID: content, a reason to enter, and feedback the system can learn from must be there to receive the user.Original episode
The distribution interface changes both viewing efficiency and creator returns. Comparing single-column and two-column layouts requires more than a single measure of viewing time. It should include ordinary creators' first feedback, continued creation, supply diversity, and user choice. Creators serve consumers both inside and outside an experiment, so ordinary user-level A/B tests can have network spillovers. A short-term consumption advantage may not mean a better long-term ecosystem. Clear segmentation, longer observation windows, and potential contamination need to be addressed; measurement difficulty does not exempt a mechanism from evaluation.
The same history contains examples of distorted proxy metrics. Red packets could raise OKRs for adding friends and interacting without creating organic relationship contexts. Changing an activity's objective near Chinese New Year might leave it delivering neither social relationships nor effective acquisition. These examples suggest that follows, shares, messages, and generation consumption are only process measures; evaluation must return to why users come back. Complaints about insufficient energy in NieTa may indicate stronger demand, but actual purchases, repeat purchases, and effects on supply still need validation. Complaints or consumption alone do not establish monetization success.
Distribution can also damage the experience awaiting users. Cheap traffic may cause registrations to surge while bringing mismatched users, overwhelming customer support, and leaving creators without feedback. Short-term CAC then looks good as subsequent value deteriorates. Growth quality should include the share of poorly matched users, task failures, complaints, refunds, negative reactions, and supply churn. “It generated exposure” is not enough to justify expenditure.
6. Growth Comes Full Circle Only When Experiment Results Feed Back into Product and Business Decisions
A growth loop has at least five steps: identify the current constraint → state the mechanism and metrics → experiment with a limited budget → examine outcomes and guardrails → decide whether to retain, revise, or stop. Mercor's description of fast decisions within a week depends on first stating what should become visible during that week and what to do if it does not. Speed does not mean omitting expectations, accountability, and failure plans. Complex or irreversible actions require higher thresholds and should not copy the pace of reversible experiments.
Opus connected customer-support issues to product priorities, fixes, and support follow-up. Community wishes, paid-service failures, and roadmap input can be collected through different channels, but someone must ultimately own closure. Automatic classification or sending an email does not mean the problem is fixed. Renewals also require separating voluntary cancellations from Past Due payment failures. Its reported improvement in retry success rates from 20%–30% to 40%–50% is experience with no stated measurement definition, not a growth-budget assumption. In practice, teams should examine eligible retries, recovered amounts, authorization, duplicate charges, and complaints, while respecting users who have already canceled.
The Gaotu episode provides an important correction to metric definitions: the official chapter heading referred to a three- or fourfold sales-conversion rate, while the detailed transcript discussed a three- or fourfold increase in “per-lead efficiency,” meaning GMV per lead. It separately described conversion approximately doubling, and only in a small-scale experiment. Business readback should prioritize specific definitions and retain the source, experimental share, baseline, and window. A chapter heading must not be inflated into a company-wide doubling of revenue. Covering more leads and answering more messages remain process counts.
7. Applying This to Product Teams: Start with a Growth Experiment Card Whose Results Can Be Verified (Editorial Proposal)
For AI companion, character, and creative products, start by fixing one region and language, one target user type, one segment of the core experience, and one channel. Avoid changing pricing, entry points, content, and recommendations in the same week. The card should record at least: the target audience and its authenticity; existing alternatives; the first action that delivers value; channel and creative versions; costs included; the experimental denominator and time window; returning use and payment; refund and quality guardrails; creator feedback; and conditions for stopping or scaling. Qualitative quotations should trace back to the corresponding participants, and behavioral outcomes to events and payment records. Surveys, demonstrations, saved admin settings, deployment, and users' actual experience must remain distinct layers of evidence.
For creator growth, first track “recruited → completes a work → receives the first meaningful consumption feedback → continues creating.” For consumer growth, track “views a card → starts interacting → receives the expected experience → returns.” Monetization then tracks actual purchases, renewals, and gross profit. Report character, category, and platform results separately. Short-term payment gains cannot offset damage to the relationship experience. Once a channel produces good results, expand audiences and scale gradually, checking each time whether the original mechanisms of trust and supply still hold. A matrix amplifies a validated mechanism; it cannot replace the mechanism itself.