Life人生
Experience is the goal; software is the current medium.体验是目的，软件是现阶段最熟悉的媒介。
Founder profile返回创始人介绍PUBLIC EDITION · 2026
FOUNDER OPERATING SYSTEM / 创始人操作系统
Make a few things
worth keeping.把少数值得存在的东西，
长期做对。
Product corrector产品修正者
Not a manifesto for making more. A handbook for deciding what deserves to exist, where it should stop and what kind of life success should leave behind.这不是一份鼓励做得更多的宣言，而是一份判断什么值得存在、产品应该停在哪里，以及成功之后应该留下怎样生活的手册。
INTENTRESULT
A small domain.
Broad coverage.
A deliberate stop.小领域。大覆盖。
以及主动停止。
The right product covers 90–95% of the core audience’s task and becomes their default entry point. It does not become a platform simply because adjacent features exist.合适的产品在边界清晰的任务域内覆盖目标用户 90%～95% 的主流需求，成为默认入口；它不会因为相邻功能存在，就把自己扩张成平台。
Product产品
Function first. Taste determines whether the work feels complete.功能优先，审美决定作品是否真正完成。
Business商业
Self-serve distribution, fair pricing and product leverage.自助分发、公平收费与产品杠杆。
Freedom自由
Success must add choice instead of building a permanent trap.成功应该增加选择权，而不是制造无法退出的系统。
01 / PRODUCT PRINCIPLES
Ten decisions that protect the product.十个保护产品的决定。
The product changes when the task changes—not when a roadmap needs filling.产品在任务发生变化时更新，而不是在路线图需要被填满时更新。
- 01
A product should change only as fast as its task domain changes.产品变化速度服从任务域变化速度，不为更新而更新。
- 02
One product should own one task in the user’s mind.一个产品只占领一个任务心智，功能不能把用户带离核心意图。
- 03
Cover 90–95% of the core audience’s tasks, then stop.对目标用户足够完整，不对所有人全能；覆盖 90%～95% 主流任务后主动停止。
- 04
Measure value as friction × frequency × people affected.用总摩擦衡量价值：单次摩擦 × 发生频率 × 用户数量。
- 05
Shorten the path from intent to result.尽量缩短意图到结果；步骤、等待、学习、上传和切换工具都属于成本。
- 06
Be honest about privacy, locality and reliability.隐私、本地化和可靠性必须诚实；端侧能完成的，不以 AI 之名强迫上传。
- 07
Price as closely as possible to how value is created.收费单位尽量接近价值产生单位。
- 08
Keep experiments separate from finished work.实验与作品分开；可以测试许多想法，但不把实验伪装成作品。
- 09
Distribution is part of the product.分发是产品的一部分；好产品不会自动被发现。
- 10
A moat means remaining hard to replace, not impossible to copy.壁垒不是别人做不出来，而是做出来以后仍然难以替代。
MOAT
Trust → habit → taste → reputation → local privacy → word of mouth → workflow品牌信任 → 使用习惯 → 设计品味 → 创始人信誉 → 隐私本地化 → 社区口碑 → 快捷工作流
02 / DECISION WHITEBOARDEliminate first.
Eliminate first.
Score second.先淘汰，
再打分。
A hard gate cannot be cancelled by an attractive score.硬性门槛不能被漂亮的总分抵消。
Hard gates硬性淘汰条件
- 1
The user, moment and reason to open it are unclear.说不清哪一群人、在什么时刻、为什么打开它。
- 2
There is no evidence of a real cost paid today.找不到用户正在支付的钱、时间、学习、错误、隐私或焦虑成本。
- 3
No market exists and there is no credible why now.市场从未成立，又说不清为什么今天与过去不同。
- 4
A small team cannot become meaningfully better in a bounded period.小团队无法在有限周期内做到明显更好。
- 5
Success requires permanent heavy sales, custom delivery or a large organization.成功必然要求长期重销售、定制、关系维护或大型组织管理。
- 6
Growth depends on endless adjacent features.必须依靠持续加入相邻功能才能增长。
- 7
No pricing model is both fair and sustainable.找不到对用户公平、对产品可持续的收费方式。
- 8
High-stakes decisions lack authoritative data, auditability and clear responsibility.高风险判断无法建立权威数据、审计追溯、责任边界和合规路径。
- 9
Even success would permanently damage freedom, craft or exitability.即使成功，也会长期损害自由、作品感或退出能力。
100
Evidence-weighted score证据加权总分
Intuition alone never earns more than the middle.“我觉得”最高只能得到中间分。
Real cost & intensity真实成本与问题强度15
Friction & time returned总摩擦与时间返还10
Solution gap & timing现有方案缺口与时机10
Founder fit创始人适配与持续意愿10
Task clarity & boundaries单一任务心智与边界10
Small-team battlefield小团队战场尺寸10
Self-serve distribution自助分发与自然增长10
Fair, sustainable pricing商业模式公平且可持续8
Freedom & exitability自由、退出与低本人依赖10
Resistance to replacement长期替代阻力7
03 / THREE DIRECTIONS
Principles, tested against real work.原则，要在真实产品里接受检验。
These are examples of three different product trade-offs, not a public roadmap.这是三种不同取舍的示例，不是对外发布的产品路线图。
SkyConvert
Validate distribution first优先验证分发闭环The clearest task promise: one place for ordinary people to convert common files. The boundary is everyday formats, not every professional edge case.任务心智最清楚：普通人需要转换常见文件时，只打开这一个工具。边界不是承诺所有专业格式，而是覆盖主流的日常转换任务。
Memora
Long-term signature candidate长期代表作候选The strongest long-term founder fit: find photos locally with words or another image. The temporary box is an action layer after search, not a complete photo-management platform.最符合长期兴趣：不用上传照片，用文字或一张图片找回相册内容。暂存盒是找到后的行动层，不扩张为完整相册管理平台。
SkyInsure
High responsibility, narrow scope高责任，谨慎收缩Narrowed from recommendation to clause understanding, coverage organization, evidence tracing and questions for qualified professionals.从替用户推荐收缩为帮助理解条款、整理已有保障、追溯证据并生成待确认的问题，把最终判断留给用户和合格专业人员。
04 / AI CONVERSATION TRACE
Not written in one pass.它不是一次写出来的。
The handbook came from an AI-led interview that kept questioning, challenging and testing the answers against real products. Exact financial figures, private work history, third-party details and emotionally charged wording were removed from this public edition.这份手册来自一场由 AI 主持、持续追问、反驳，并用真实产品校验答案的访谈。公开版隐去了精确财务数字、私人合作经历、第三方细节和情绪化原话。
Read the shared conversation阅读公开对话原文01Separate the life goal先分清人生终局
AIHow do craft, wealth, influence and freedom really rank?作品、财富、影响力与自由，真正的排序是什么？
CREEPONSKYFreedom comes first, then craft. Wealth is a threshold that buys choice; influence is not an active goal.自由排在最前，作品其次。财富是一条购买选择权的安全线，影响力不是主动目标。
02Identify the creative reward识别创作快感
AIIs the reward novelty, business discovery, or turning ten steps into one?更兴奋的是技术突破、商业发现，还是把十步变成一步？
CREEPONSKYThe durable reward comes from correcting clumsy flows while making function and taste cohere.稳定的快感来自修正笨拙流程，并把功能与审美一起做到成立。
03Map the problem radar找到问题雷达
AIWhere does “why can’t this work better?” appear most often?在哪些领域最容易冒出“为什么不能这样做”？
CREEPONSKYPhotos, files, search, conversion, automation, system tools and personal digital assets.照片、文件、搜索、转换、自动化、系统工具与个人数字资产。
04Define the task boundary定义任务边界
AIOne feature, or a whole task domain?产品应该窄到只有一个功能，还是覆盖整个任务域？
CREEPONSKYA single mental category: a small domain with broad coverage, stopping before professional edge cases.不是单功能，而是单一心智：小领域，大覆盖；专业边缘需求交给专业工具。
05Rewrite the moat重写壁垒
AIIf feature and UI are copied, why stay?功能和 UI 都被复制以后，用户为什么不走？
CREEPONSKYTrust, habits, taste, reputation, local privacy, word of mouth and workflow integration.品牌信任、习惯、设计品味、信誉、隐私本地化、口碑和工作流嵌入。
06Put distribution back into product把分发放回产品
AIWould SEO, content, funnels and pricing still feel like product work?是否愿意用一年研究 SEO、内容、漏斗、价格与用户流失？
CREEPONSKYYes. The objection is relationship-led growth, not measurable product-led growth.愿意。排斥的不是增长，而是依赖关系与长期商务维护的增长。
07Test against real products用真实产品校验
AIWhich direction fits today, and which looks like the long-term work?三个已有方向，哪一个适合当下，哪一个更像长期作品？
CREEPONSKYSkyConvert tests distribution first; Memora keeps signature potential; SkyInsure narrows around responsibility.SkyConvert 先验证分发闭环；Memora 保留代表作潜力；SkyInsure 因责任边界而收缩。
05 / BEFORE BUILDING
Twelve questions before code.写代码前，先回答十二个问题。
If the answers depend on imagination, the project is not ready.如果回答依赖想象，项目还没有进入正式开发的资格。
- 01
Who must complete what task, and at what moment?这是谁在什么时刻必须完成的什么任务？
- 02
How is it solved today, and what does that cost?他今天如何解决，正在支付什么成本？
- 03
Is it intense pain, or small friction at high frequency?它是强痛点，还是弱痛点但极高频？
- 04
Why do current products fail, and why can today be different?为什么现有产品没解决好？为什么今天能做得更好？
- 05
What one sentence will users remember?用户会用哪一句话记住它？
- 06
What must be built, and what must be refused?为覆盖主流需求，必须做什么；坚决不做什么？
- 07
Can a small team become meaningfully better in a bounded period?小团队能否在有限周期内做到明显更好？
- 08
How can users find, understand and buy it unaided?用户如何独立找到、理解和购买它？
- 09
Does pricing match how value is created?收费是否接近价值产生方式？
- 10
Why stay after the feature is copied?功能被复制后，用户为什么仍不愿意离开？
- 11
Would success create freedom or lock the founder in?产品成功后，我会更自由，还是被锁住？
- 12
What evidence remains after failure, and can daily operations be exited after success?失败能留下什么证据；成功能否退出日常运营？
Clear boundaries. Fair pricing. No feature pollution. Built to last. Fast where speed matters.功能边界清楚。不乱收费。不乱加东西。用得久。该快的地方足够快。
Back to the founder profile返回创始人介绍