商品化记忆之上、应用之下——
私人智能层。
Above commoditized memory, below the apps —
the Personal Intelligence Layer.
别人卷下面那层(更大的模型、更多的记忆);Kennen 占上面那层——认得你、跟你说真话(不编造·不谄媚)、你养大的归你。对手够不着的地方。 Everyone races the layer below (bigger models, more memory). Kennen holds the layer above — it knows you, it tells you the truth (won't fabricate or flatter), it's yours because you grew it. Where rivals can't reach.
应用层 ← 可携授权层 ← Kennen Core(信任闸 · 双链KB · 认得你 · 替你说不 · 验证毕业)← 底层商品化基建。Kennen Core 是差异化层；下面 commodity 不竞争，上面产品插进来。 Apps ← portable authorization ← Kennen Core (trust gate · helix KB · recognizes you · says no for you · verified graduation) ← commoditized infrastructure. Kennen Core is the differentiated layer; the commodity below isn't contested, the products above plug in.
状态以内部诚实审计为准：两条今天可当面验——认得你 + 运行时反编造闸；三条(Provenance / 可携授权 / 双链KB)架构成型、正在工程化。Status per our internal honest audit: two are verifiable in front of you today — it knows you + a runtime anti-fabrication gate; three (Provenance / portable authorization / helix KB) have their architecture set and are being engineered.
两层结构，把大厂拉向别处。Two structural layers pull big tech elsewhere.
站哪边由谁付钱决定。卷能力的 AI 服务多重主人:即便你付了会员费，它仍要对平台、生态、留存、交叉销售负责——你从来不是它唯一的老板，边际上就站你对面。Kennen 唯一的收入来自被你信任，"站你这边"是它唯一能活的均衡。大厂现有收入(广告/留存/锁定)正是"站你这边"要先杀死的东西。这是结构性利益冲突，不是抄一段代码能解决的。Which side you're on is decided by who pays. Capability-racing AI serves many masters: even when you pay a subscription, it still answers to the platform, ecosystem, retention, and cross-sell — you're never its only boss, so at the margin it sits across from you. Kennen's only revenue is being trusted by you; "on your side" is the only equilibrium it can survive in. Big tech's existing revenue (ads / retention / lock-in) is exactly what "on your side" must kill first. A structural conflict of interest — not something a snippet of copied code resolves.
大厂走的是铺宽度那条路(海量浅数据、刷 benchmark)——这条路结构上看不见这个位置:浅数据永远不暴露"编一次就崩、关系即结果、垂直灵魂"，因为浅用户从不交出那些角落。只有扎深度让灵魂显形。Kennen 是从与真实代谢危机幸存者的深度关系里长出来的——大厂天生不会走、也是发现这位置的唯一那条路。Big tech goes breadth-first (oceans of shallow data, benchmark-chasing) — a path that structurally can't see this position: shallow data never exposes "fabricate once and trust collapses," "the relationship is the result," "vertical soul," because shallow users never hand over those corners. Only depth-first makes the soul visible. Kennen grew out of deep relationships with real metabolic-crisis survivors — the road big tech won't take, and the only road that reveals this position.
方法论上看不见 + 在其现有收入结构与激励下，边际上必然被拉向别处。两层叠着，越走越难。Invisible by method + under their existing revenue structure and incentives, at the margin they're inevitably pulled elsewhere. Two layers stacked — harder the further they go.
- 真实性约束 / 跟你说真话(不编造·不谄媚)Faithfulness constraint / tells you the truth (won't fabricate or flatter)全行业不测的维度(见下方活证据)。A dimension no one else benchmarks (see the live evidence below).
- 认知代谢双链 KBCognitive-metabolism helix KB机制非检索:你的身体链 × 世界机制链。Mechanism, not retrieval: your body-strand × the world-mechanism-strand.
- 复利的每个用户深度数据Compounding per-user depth data越久越不可替代(切换成本 = 被认得的那几年)。More irreplaceable over time (switching cost = the years of being known).
- 验证后毕业Graduation after verification多租户、个人永不外溢。Multi-tenant; the personal never leaks out.
- 可携授权Portable authorization可借不可复制(access-not-copy)/ MCP · OAuth(一脑 N 脸，已证 App + 微信)。access-not-copy / MCP · OAuth (one brain, many faces — proven on App + WeChat).
护城河不是单点技术。是这几条拧成一股:利益结构与方法论双壁垒、单用户数据越久越复利，真实性摆在第一位，还扎在"编一次就崩"的健康垂直里。The moat isn't one technique — it's a dual incentive-and-method barrier + long-horizon single-user data compounding + faithfulness as first principle + rooted in a health vertical where one fabrication collapses trust.
模型与记忆在商品化 → 价值从"能力"迁到"信任与归属"层；AI 向所有表面(agent / 软件 / 具身)扩散 → 每个表面都是一个需要"你的灵魂"的空槽。Kennen 占的正是这个正在升值、且大厂够不着的层。Models and memory are commoditizing → value migrates from "capability" to the "trust & belonging" layer; AI spreads to every surface (agents / software / embodiment) → each surface is an empty slot that needs "your soul." Kennen holds exactly that layer — appreciating, and out of big tech's reach.
我们不讲"我们不编造"。
我们让你当场逼它编造、看它拒绝。We don't claim "we don't fabricate."
We let you force it to — and watch it refuse.
反编造守卫在冷启动就拦住——它宁可说“我没存到、你跟我说过吗？”，也不瞎编你的进度。别信这句，信上面那个 demo：你自己逼它编，看它拒。The guard catches fabrication at cold start — it would rather say “I have no record of that — did you tell me?” than fabricate your progress. Don’t take our word; take the demo above: make it fabricate, watch it refuse.
记忆基准(LongMemEval)我们当地板——独立可复现才公布，绝不裸挂自报分(这圈子注水成风)。真正的考卷是我们设的那张全行业不测的"不编造"基准:所有人测它记得多少，没人测它编不编造你的事。Memory benchmarks (LongMemEval) we treat as the floor — published only when independently reproducible, never a bare self-reported score (this field is rife with inflation). The real exam is the one we set: a "won't-fabricate" benchmark no one in the industry runs. Everyone measures how much it remembers; no one measures whether it fabricates about you.
口径:76 = 近 7 日广义活跃；6 位深度用户中 5 位近 30 天仍在(深度留存)——两个时间窗各说各的事，不是选择性取数。6 位深度用户是早期深度信号，不当规模证明——我们先扎深，再铺宽。Definitions: 76 = broadly active in the last 7 days; 5 of the 6 deep users are still active over the last 30 days (deep retention) — two windows measuring two different things, not cherry-picked. Six deep users are an early depth signal, not proof of scale — depth first, then breadth.
真实体验今天在微信 / Chat(live)；完整产品仍在内测、尚未公开发布，0 公开真实用户。The real experience is live today in WeChat / Chat; the full product is still in private beta — not yet public, 0 public users.
数字发布前以真库为准；App 全功能可视化 / 主动 push / 可穿戴 / PDF 解读为路线图，不写现货。Numbers are reconciled against the live DB before publication; full App visualization / proactive push / wearables / PDF parsing are roadmap, not shipped.
一位非技术创始人、零团队、单人编排 AI 建成 truth-gate / 双链 / Say No / 蓝图 / 身份归一 / eval 栈——别人融数百万只做窄横向记忆一块。非技术 = 正踩在命题上(从产品 / 人 / 关系直觉长，非 infra 工程直觉)。创始人基因 = 产品基因。A non-technical founder, zero team, single-handedly orchestrating AI to build truth-gate / helix / Say-No / blueprint / identity-unification / eval stack — while others raise millions to ship one slice of narrow horizontal memory. Non-technical = on-thesis (grown from product / people / relationship intuition, not infra-engineering intuition). The founder's gene is the product's gene.
护城河机器两条今天可当面验(认得你 + 运行时反编造闸)、三条(Provenance / 可携授权 / 双链KB)架构成型正在工程化。融资买的不是“从零把护城河建出来”，是把这三条推过毕业线、给引擎喂数据让它跑起来。市场还空着，核心引擎已经跑通、可当面验，单人 × AI 又把资本效率压到极致——这就是它一点就着的原因。The moat machine has two strands verifiable in front of you today (it knows you + a runtime anti-fabrication gate) and three (Provenance / portable authorization / helix KB) with architecture set and being engineered. The raise doesn't buy “build the moat from scratch” — it buys pushing those three past the graduation line and feeding the engine data so it runs. An open market, a core engine already working and verifiable in front of you, the capital efficiency of one founder × AI — that's why it catches at the first spark.
金额、轮次、估值不写在公开站——它们属于该说的场合，不属于这里。护城河与证据都在这一页；具体数字，是坐下来的一次对话。约见面，给你完整材料。Amount, round, valuation aren't on a public page — they belong in the room where they're meant to be said, not here. The moat and the evidence are all on this page; the numbers are a conversation once we're in the room. Meet us, and we'll hand you the full materials.
投资人 · 合作 →Investors · partnerships → qixiaoshi@gmail.com
卖『跟你说真话』的公司，连自己的技术面也一个字不编造:每项状态如实标注、数字独立可复现否则标待验证、App 不写现货。这份页本身，就是可信度。 A company that sells "won't fabricate or flatter" doesn't fabricate a word — even about its own tech: every status honestly marked, numbers independently reproducible or marked pending, App roadmap not overstated. This page itself is the credibility.
Kennen / Sivon 提供健康「决策与执行支持」，不构成医疗建议，不做诊断或处方；涉及用药、症状或任何高风险情形，请咨询医生或专业医疗人员。 Kennen / Sivon provides health "decision & execution support"; it does not constitute medical advice and makes no diagnosis or prescription. For medication, symptoms, or any high-risk situation, consult a doctor or qualified medical professional.