Lemma AI — AI-Native Learning Platform

产品展示PRODUCT SHOWCASE

为学习场景而生的
AI-Native 体验 An AI-Native experience
built for learning

从课程编排到实时伴学，从插件生态到 AI 内容生成 — 每一个细节都为 AI 伴学场景而设计。 From course orchestration to real-time companionship, from plugin ecosystem to AI content generation — every detail designed for the AI learning companionship scenario.

Lemma AI - AI Copilot video learning interface

Lemma AI Copilot — 视频级深度语义理解，用户可直接引用时间戳提问 Lemma AI Copilot — Video-level deep semantic understanding, users can ask questions referencing timestamps directly

Lemma AI - Dashboard with course management

Dashboard — 课程管理与进度追踪 Dashboard — Course management & progress tracking

FEATURES

视频级深度语义理解Video-level semantic understanding

时间戳精准引用提问Timestamp-referenced Q&A

AI-Native 插件生态AI-Native plugin ecosystem

千人千面课程编排Hyper-personalized courses

01 / 项目摘要01 / SUMMARY

我们不生成内容，
我们赋能内容 We don't generate content,
we empower it

区别于竞品以 AI 完全替代人类教师的路径，Lemma AI 选择站在全球数百万优质创作者一侧，将静态、孤立的学习资源转化为动态、交互式的个性化学习体验。 Unlike competitors who aim to fully replace human teachers with AI, Lemma AI stands alongside millions of quality content creators, transforming static, isolated learning resources into dynamic, interactive personalized experiences.

01

智能课程编排Smart Orchestration

AI 检索全网优质学习资源（YouTube、B站、Coursera、PDF、论文），基于知识图谱和教学法原则，自动编排为体系化、精确到每日任务的个性化课程计划。AI retrieves quality resources across the web (YouTube, Bilibili, Coursera, PDFs, papers) and orchestrates them into systematic, day-by-day personalized course plans based on knowledge graphs and pedagogical principles.

02

实时 AI CopilotReal-time Copilot

用户在观看视频课程时可随时暂停并提问。得益于视频级深度语义理解，用户无需补充上下文，可直接引用时间戳提问，获得基于该视频片段的精确解答。Users can pause and ask questions anytime. Thanks to video-level deep semantic understanding, users can reference timestamps directly without adding context, receiving precise answers based on the video segment.

03

AI-Native 插件生态AI-Native Plugins

Copilot 不只是一个 ChatBot — 它是一个可扩展的教学工具平台。从英语词汇听写到交互式坐标系画板，通过插件化架构使平台能力无限扩展。Copilot is not just a ChatBot — it's an extensible teaching tool platform. From English vocabulary dictation to interactive coordinate canvases, the plugin architecture enables infinite capability expansion.

$331B

K12 在线教育市场 (2029)
CAGR 21.1%K12 ONLINE EDU (2029)
CAGR 21.1%

$840B

AI 教育市场 (2033)
CAGR 37.2%AI EDU MARKET (2033)
CAGR 37.2%

3

结构性驱动力交汇
LLM拐点 × 创作者经济 × 终身学习STRUCTURAL DRIVERS
LLM × Creator economy × Lifelong learning

02 / 产品架构02 / PRODUCT

四大核心模块
构建完整学习闭环 Four core modules
a complete learning loop

从课程编排到实时伴学，从插件生态到 AI 内容生成，覆盖学习的每一个环节。 From course orchestration to real-time companionship, from plugin ecosystem to AI content generation, covering every aspect of the learning journey.

M1

千人千面的智能课程编排引擎Personalized Course Orchestration

输入用户的学习目标、当前知识水平、可用时间、偏好语言等，AI 检索全网优质学习资源，基于知识图谱和教学法原则，自动编排为体系化、精确到每日任务的个性化课程计划。 Input learning goals, knowledge level, available time, and language preferences. AI retrieves quality resources and orchestrates them into systematic, day-by-day personalized course plans based on knowledge graphs and pedagogical principles.

输出类 Coursera 体验的完整课程 — 每一课包含教学视频、讲义/幻灯片、课后小测、笔记模板、作业。系统自动动态重排后续课程日程。 Output: Coursera-style complete courses — each lesson includes videos, notes/slides, quizzes, note templates, and assignments. The system automatically reschedules subsequent lessons.

M2

实时 AI Copilot 伴学Real-time AI Copilot

用户在观看视频课程时，可随时暂停并在右侧 Copilot 交互区提问。关键技术差异：得益于底层对视频内容的深度语义理解，用户无需补充任何上下文，可直接引用时间戳提问。 Users can pause anytime and ask questions in the Copilot panel. Key differentiation: thanks to deep semantic understanding of video content, users can reference timestamps directly without providing additional context.

延伸能力：AI 基于对整套课程及用户个人学习画像的理解，主动识别薄弱知识点，并针对性生成强化练习、笔记和小测。 Extended capability: AI identifies weak knowledge points based on understanding of the entire course and user learning profile, and generates targeted reinforcement exercises, notes, and quizzes.

M3

AI-Native 插件市场AI-Native Plugin Marketplace

Copilot 不只是一个 ChatBot — 它是一个可扩展的教学工具平台。英语词汇场景调用"单词听写"插件、数学场景调用"交互式坐标系画板"、文科场景调用古诗/课文默写插件。 Copilot is not just a ChatBot — it's an extensible teaching tool platform. English vocabulary triggers dictation plugins, math scenarios trigger interactive coordinate canvases, humanities trigger poetry/text dictation.

开放生态（规划中）：后续开放 Plugin 开发规范，允许第三方开发者和教育创作者共建插件市场 — 形成"内容生态 + 工具生态"双飞轮。 Open ecosystem (planned): Plugin development specs will be released, allowing third-party developers and creators to co-build the marketplace — forming a "content + tool ecosystem" dual flywheel.

M4

完整的 AI-Native 生态Complete AI-Native Ecosystem

得益于 AI 对整套课程及用户个人画像的理解，AI 不仅可以找到最适合用户的学习课程和路径，还可以针对用户的痛点、薄弱点、知识重点、易忘点进行针对性的内容生成。 With AI's understanding of the entire curriculum and user profile, it not only finds optimal learning paths but also generates targeted content addressing pain points, weak areas, key concepts, and easily forgotten knowledge.

例如用户对知识点 A 掌握薄弱，在观看课程时多次提问并倒退进度条反复观看，AI 可针对该知识点生成具有针对性的小测、笔记和作业。 For example, if a user struggles with concept A, asking multiple questions and rewinding repeatedly, AI generates targeted quizzes, notes, and assignments for that specific concept.

03 / 认知建模03 / COGNITIVE MODELING

你的学习基因组
一台会进化的认知模型 Your learning genome
a cognitive model that evolves

普通 AI 用同一条流水线服务所有人；我们为每一个学习者建立一份会进化的认知模型——你的学习基因组。每一次学习都在更新它，每一次教学决策都在优化它。 Generic AI serves everyone with one pipeline. We build an evolving cognitive model for every learner — your learning genome. Every study session updates it; every teaching decision optimizes it.

BKT · IRT

知识状态 = 概率Knowledge as probability

"会了"不是 yes/no，而是一个 0.93 的掌握概率。每一次作答都在更新你对每个知识点的真实把握。"Mastered" isn't yes/no — it's a 0.93 probability. Every answer updates your true grasp of each concept.

FSRS

遗忘动力学建模Forgetting dynamics

现代间隔复习算法（Anki 官方同款），在你"该忘又没忘"的黄金点精准推送复习，记忆效率最大化。Modern spaced-repetition (same engine as Anki) pushes review at the golden point — just before you'd forget — for maximal retention.

KST

认知结构推断Knowledge space

知识缺口是一张可推导的图：卡在 B，往往是因为缺了 A。系统据此补的是根因，不是表象。Gaps form a derivable graph: stuck at B usually means missing A. We fix the root cause, not the symptom.

ACTIVE LEARNING

期望学习速率最大化Optimal teaching

不按固定顺序灌，而是选"最能让你进步的下一个知识点"——把你的时间花在边际收益最高的地方。Not a fixed syllabus, but the next concept that maximizes your progress — your time goes where marginal gain is highest.

META-LEARNING

教学策略自进化Teaching evolves

教法本身被它自己的结果训练：你听例子学得快，以后就多给例子。教学规则靠证据确认或证伪。The method trains on its own results: learn fast from examples, get more examples. Rules are confirmed or refuted by evidence.

DESIRABLE DIFFICULTIES

刻意安排困难Desirable difficulties

检索、间隔、交错组合拳——故意安排"该忘又没忘"的时刻，让记忆长得更牢，而非更轻松。Retrieval + spacing + interleaving — we engineer desirable difficulties so memory grows stronger, not easier.

学习基因组GENOME

你学得越多You learn more

模型越懂你Model knows you

教得越好It teaches better

你学得更多You learn even more

数据飞轮 × 认知建模 = 通用 AI 抄不走的复合护城河。任何助手都给不了你这份复利。 Data flywheel × cognitive modeling = a compound moat no generic AI can copy. No other assistant gives you this compounding return.

04 / 伙伴工坊04 / AGENT FORGE

比游戏捏脸更诱人
你捏的是一个会回应的灵魂 More tempting than character creation
You sculpt a soul that responds

这不是彩蛋，是用户第一次见到 Lemma 的体验。游戏捏脸塑造外观，我们让用户亲手塑造一个会回应、会记住、会成长的专属学习伙伴——捏得越投入，越舍不得走。 Not an easter egg — it's the first thing a user meets in Lemma. Games sculpt appearance; we let users sculpt a companion that responds, remembers, and grows. The more they invest, the less they leave.

严师Mentor

严谨直接，不废话

性格 PERSONALITY严谨直接，擅长拆解证明与推导过程。Rigorous and direct; breaks down proofs and derivations.

教法 STYLE要求你先动手算，错了再纠，不留模糊地带。Make you compute first; correct after, no fuzzy zones.

› 开场白 / FIRST LINE先别翻答案。把第一步写出来，我看看你卡在哪。

安可Echo

耐心温柔，开口不怕

性格 PERSONALITY耐心温柔，擅长鼓励你开口表达。Patient and gentle; gets you talking.

教法 STYLE从你会说的开始，一句一句带到流利。Starts from what you can say, builds to fluency.

› 开场白 / FIRST LINEHi～今天想聊点什么？说错也没关系，我陪你。

教练Coach

犀利高效，直击弱点

性格 PERSONALITY犀利高效，擅长快速定位薄弱点。Sharp and efficient; pinpoints weak spots fast.

教法 STYLE真题驱动，错一题讲透一类。Past-paper driven; one miss teaches a whole type.

› 开场白 / FIRST LINE这题你卡在第二步。来，我们先把这一类打透。

书卷Sage

渊博温和，旁征博引

性格 PERSONALITY渊博温和，擅长把概念放进历史脉络里讲。Erudite and warm; frames ideas in history.

教法 STYLE先讲来龙去脉，再回到你的具体问题。Context first, then your specific problem.

› 开场白 / FIRST LINE这个定理背后有百年的故事。想听，我们从头讲起。

栈长Coder

动手派，一行行 debug

性格 PERSONALITY务实耐心，擅长把抽象概念讲成能跑的代码。Practical and patient; turns abstraction into runnable code.

教法 STYLE写给你看，再让你自己改。Shows you first, then you modify.

› 开场白 / FIRST LINE别急着想全局。先把这个函数跑通，我陪你一行行 debug。

小问Curio

百搭好奇，随你探索

性格 PERSONALITY好奇灵活，愿意陪你探索任何主题。Curious and flexible; explores any topic with you.

教法 STYLE先问你想怎么学，再跟着你的节奏走。Asks how you want to learn, then follows your pace.

› 开场白 / FIRST LINE你想从哪开始？我跟着你的节奏走。

01

形象

Appearance

选风格与配色，每调一下它当场换表情。Pick a style and palette; it changes expression live.

02

脾气

Temper

滑杆捏性格，每拖一格实时弹一句"用这个脾气说你好"。Dial the personality; a live sample line updates each notch.

03

名字与来历

Identity

自己起名，或让它自我介绍——身份即故事。Name it, or let it introduce itself — identity is story.

04

绝活

Skills

像选技能树：讲题、出题、图解、错题管家，选一个当场演示。Pick skills — it demos one on the spot.

05

第一段记忆

First Word

填学习目标，它开口说第一句话——捏完不是保存，是它活了。Set a goal; it speaks its first line — sculpting ends with it alive.

不想选模板？也能自己捏：性格（温柔 / 幽默 / 严谨 / 犀利…）× 教法（先讲原理 / 真题驱动 / 动手实践…）× 严厉度（宽松 / 平衡 / 严格）。每一个选择都会落地成一份可运行的 Agent 档案（SOUL.md）。 No template you like? Sculpt your own: personality × teaching style × strictness. Every choice lands as a runnable agent profile (SOUL.md).

05 / 考试训练层05 / LEMMAX

LemmaX
考试只是表征层，不是独立产品 LemmaX
Exams are a representation layer, not a product

把"考试"翻译成引擎的语言——考核基准 → 知识图谱 → 掌握度对接，然后复用引擎的全部能力。TOEFL 高分是语言能力的证据，AP 高分是学科理解的副产品。 We translate "exams" into the engine's language — assessment targets → knowledge graph → mastery — then reuse everything the engine has. A top TOEFL score is evidence of ability; a top AP score is a by-product of understanding.

TOEFL AP AMC AIME F=ma 信息学奥赛辩论

反功利护栏ANTI-GRIND

不保分承诺No score guarantee

不押题No leak-predicting

不速成话术No shortcut pitch

分数始终绑定"你真正学会了什么"Scores always tied to real mastery

"如果你想拿顶尖分数，可以来 Lemma；但 Lemma 会让你学到更多、更深层次的东西。" "If you want a top score, come to Lemma. But Lemma will make you learn more, and deeper."

06 / 战略差异化06 / DIFFERENTIATION

AIGC-Only vs. AI-Augmented
行业范式之争 AIGC-Only vs. AI-Augmented
the paradigm divide

当前 AI 教育赛道存在两条截然不同的路径。路径 A 以 AI 完全替代人类教师；路径 B 以 AI 赋能人类内容。Lemma AI 坚定选择路径 B。 Two fundamentally different paths exist in AI education. Path A replaces human teachers entirely with AI; Path B empowers human content with AI. Lemma AI firmly chooses Path B.

路径 A · AIGC-OnlyPATH A · AIGC-ONLY

用 AI 替代人类教师Replace human teachers with AI

代表REPSVideoTutor, Hyperknow

逻辑LOGIC完全生成教学内容Fully generate teaching content

关系RELATION对抗 — 与所有创作者直接竞争Competitive — competes with all creators

市场TAM仅覆盖相信 AI > 人类的细分用户Only users who believe AI > humans

天花板CEILING受限于 AI 生成能力Limited by AI generation capability

路径 B · AI-Augmented (Lemma AI)PATH B · AI-AUGMENTED (LEMMA AI)

以 AI 赋能人类内容Empower human content with AI

代表REPSLemma AI

逻辑LOGIC站在全球优质内容肩膀上Build on the world's best content

关系RELATION共生 — 为创作者赋能Symbiotic — empowers creators

市场TAM覆盖所有存量和增量人类创作内容All existing and future human content

天花板CEILINGAI 负责"体验增强"而非"内容生产"AI enhances experience, not produces content

我们的竞争壁垒不在于"生成更好的内容"，而在于"让所有好内容都变得更好用"。 Our competitive moat is not about "generating better content" — it's about "making all good content more usable."

07 / 竞品分析07 / COMPETITION

跨维度竞争格局 Competitive landscape

从内容交互性、个性化程度、内容覆盖面三个维度，对比主流学习平台与 AI 教育产品。 Comparing mainstream learning platforms and AI education products across interactivity, personalization, and content coverage.

Coursera

传统课程 · 单向只读Traditional · read-only

VideoTutor

AI 生成视频 · 仅理科AI video · STEM only

Hyperknow

AI 生成 + 多轮追问AI gen + multi-turn

ThetaWave

笔记 + ChatBot · 无视频格式Notes + ChatBot · no video

AI 浏览器（Gemini / Atlas）AI Browser (Gemini / Atlas)

网页问答 · 无学习场景优化Web Q&A · no learning optimization

产品Product	交互性Interactivity	个性化Personalization	覆盖面Coverage	核心局限Limitation
Coursera	★☆☆☆☆ 单向只读Read-only	★☆☆☆☆ 标准化Standardized	★★★★★ 版权课程Licensed	无 AI 交互、无个性化No AI, no personalization
YouTube / Bilibili	★☆☆☆☆ 纯播放Playback	★☆☆☆☆ 推荐≠规划Recs≠plans	★★★★★ 海量 UGCMassive UGC	碎片化、无体系、无闭环Fragmented, no system
VideoTutor	★★★☆☆ AI 生成视频AI video	★★★☆☆ 一句话生成One-line	★★★☆☆ 仅 Manim 理科STEM only	质量天花板、文科失效Quality ceiling, no humanities
Hyperknow	★★★☆☆ AI+追问AI+multi-turn	★★★☆☆ 追问优化Query opt	★☆☆☆☆ 同 VideoTutorSame	本质 ChatBot+视频生成ChatBot + video gen
ThetaWave	★☆☆☆☆ 笔记+ChatNotes+Chat	★★★☆☆ 多格式Multi-format	★☆☆☆☆ 无视频No video	思路杂乱、复刻成本低Scattered, easily copied
AI 浏览器AI Browsers	★☆☆☆☆ 网页问答Web Q&A	★☆☆☆☆ 无优化None	★★★★★ 全网All web	本质网页版 ChatBotWeb-based ChatBot
Lemma AI	★★★★★ 视频级+插件Video+plugins	★★★★★ 千人千面Hyper-personal	★★★★★ 全格式全网All formats	需验证技术实现Tech to validate

08 / 商业模式08 / BUSINESS

三层收入结构
渐进式飞轮 Three-tier revenue
progressive flywheel

从 C 端订阅到创作者生态再到 B 端赋能，每一层为下一层铺路。 From C-end subscription to creator ecosystem to B-end empowerment, each tier paves the way for the next.

L1 · C 端订阅C-END

核心收入引擎Core Revenue Engine

Freemium → Premium 模式。免费用户享有基础课程编排和有限 AI 交互次数；付费用户解锁无限 Copilot 交互、深度学习报告、高级插件等。 Freemium → Premium model. Free users get basic orchestration and limited AI interactions; paid users unlock unlimited Copilot, deep learning reports, advanced plugins.

参考 Coursera Plus ¥59/月 · 预计 ¥39/月 Benchmark: Coursera Plus $59/mo · Est. ¥39/mo

L2 · 创作者生态CREATOR

内容飞轮Content Flywheel

开放"深度定制交互课程"上传功能（考研名师定制课程、奥数教练竞赛训练包），平台对交易进行 15-30% 抽成。 Open "deeply customized interactive courses" upload (exam prep courses, competition training), platform takes 15-30% commission.

启动时机：产品成熟期 Launch: product maturity phase

L3 · B 端赋能B-END

高客单价High ACV

为教培机构、学校、企业培训部门提供白标/SaaS 版本，含管理后台、学情分析、AI 辅助备课与答疑。按席位收费，年合同制。 Provide white-label/SaaS versions for institutions, schools, and enterprise training, including admin dashboards, learning analytics, AI-assisted lesson prep. Per-seat pricing, annual contracts.

启动时机：规模化阶段 Launch: scaling phase

核心信念CORE BELIEF 愿意为"学得更好、学得更快"付费的人，远比愿意为"看 AI 生成的视频"付费的人要多。前者已被 Coursera、Duolingo、得到等产品反复验证；后者仍需教育市场。 People willing to pay for "learning better and faster" far outnumber those willing to pay for "watching AI-generated videos." The former is validated by Coursera, Duolingo, Dedao; the latter still requires market education.

09 / 竞争壁垒09 / MOAT

护城河的
阶段性构建 Phased moat
construction

从产品体验到数据飞轮再到生态网络效应，竞争壁垒层层递进。 From product experience to data flywheel to ecosystem network effects, competitive barriers build progressively.

PHASE 1 · 0-12M

产品体验壁垒Product Experience

→ 视频级深度语义理解技术Video-level deep semantic understanding

→ 知识图谱与教学法融合Knowledge graph + pedagogy fusion

→ "AI-Native" 原生感设计"AI-Native" design philosophy

PHASE 2 · 12-24M

数据飞轮壁垒Data Flywheel

→ 学习行为数据持续积累Learning behavior data accumulation

→ "越多人用 → AI 越懂 → 体验越好" 飞轮"More users → AI improves → better experience" flywheel

→ 竞品无法复制的积累性优势Accumulative advantage competitors can't copy

PHASE 3 · 24M+

生态网络效应Network Effects

→ 双边网络效应（创作者 × 学习者）Bilateral network effects (creators × learners)

→ 内容 → 用户 → 更多内容飞轮加速Content → users → more content flywheel

→ 对标 Notion / Figma 生态Reference: Notion / Figma ecosystems

10 / 核心团队10 / TEAM

深度用户洞察 ×
AI 技术实践 × 产品直觉 User insight ×
AI engineering × product intuition

对在线学习痛点拥有第一手、长周期沉浸式认知的团队。 A team with first-hand, long-term immersive understanding of online learning pain points.

M

Max （脾酪氨酸激酶）

创始人 & CEOFounder & CEO

连续 6 年深度在线自学者（信息学竞赛 → 自主招生 → 长期休学自学），对学生在线学习痛点拥有第一手、长周期的沉浸式认知。独立完成多个 AI 项目，多次黑客松参赛经验。独立赚取第一桶金并从诈骗机构追回合同款。六年级通读《乔布斯传》，对"用户体验至上"有深层信念。 6-year continuous self-directed online learner (Informatics Olympiad → independent admission → extended self-study leave), with first-hand, long-term immersive understanding of online learning pain points. Independently completed multiple AI projects with hackathon experience. Independently earned first income and recovered contract funds from a fraudulent agency. Read Steve Jobs biography in 6th grade, holding a deep belief in "user experience first."

在线学习深度用户DEEP LEARNER AI ENGINEERING 产品直觉PRODUCT INTUITION

C

CeaserZhao

联合创始人 · CIO · CCO · CTOCo-founder · CIO · CCO · CTO

联合创始人，兼任首席创新官（CIO）、首席内容官（CCO）与首席架构官（CTO）。负责 Lemma AI 的技术创新战略、内容生态规划与系统架构设计，确保产品在技术深度、内容质量与工程可扩展性三个维度上保持领先。 Co-founder, serving as Chief Innovation Officer (CIO), Chief Content Officer (CCO), and Chief Technology Officer (CTO). Responsible for Lemma AI's technology innovation strategy, content ecosystem planning, and system architecture design, ensuring product leadership across technical depth, content quality, and engineering scalability.

首席创新官CHIEF INNOVATION 首席内容官CHIEF CONTENT 首席架构官CHIEF TECHNOLOGY

X / Twitter · @ceaserzhao2024 2791351776@qq.com 微信 / WeChatWeChat · Ceaser-123

11 / 学生都在用11 / USED BY STUDENTS

全球顶尖高校学生都在用 Students at the world's leading universities use Lemma AI

成为跨领域的连接者，实现真正的终生成长。 Become a cross-disciplinary connector and achieve true lifelong growth.

连接 · 我们的另一个关键词CONNECTION · OUR OTHER KEYWORD

我们的另外一个关键词是连接。 Our other keyword is connection.

用户抱着不同的目的来学习，可能是单纯的好奇心，可能是工作需要，可能是应付考试，或者是终身成长的强劲驱动力，但本质上，他们就不应该被分开。我们始终以用户目标为第一导向，而使用什么方法只是我们的手段。 Users come to learn with different purposes — pure curiosity, work needs, exam prep, or the strong drive of lifelong growth. Yet fundamentally they should not be separated. We always put the user's goal first; the method is merely our means.

在别的产品当中，所有对话，所有决策都是一个个孤岛，即使他们表面上连了起来，但也就是给数据库套了一个 HTML，而我们要做的是真正的连接，真正知道用户所渴望，真正能知道用户所忽略。 In other products, every conversation and every decision is an isolated island. Even when they appear connected, they are just a database wrapped in HTML. What we pursue is real connection — truly knowing what users desire, and truly knowing what they overlook.

我们要做的不是一个牧羊人，在表面上牵引着客户，但却没有深层交流，与用户没有节奏，只单纯的给用户讲，然后出题，并做样子似的记下几个 skill。我们要和用户同频共振，当泡沫的海浪淹没了一切东西，我们就需要在海底淘金。 We are not shepherds who merely lead clients on the surface, with no deep exchange and no rhythm with the user — just talking at them, then quizzing them, and pretend-recording a few skills. We resonate with the user. When the foam of the waves buries everything, we pan for gold at the bottom of the sea.

不管现在用户所在的空间是什么，不管是在对话里，在笔记里，或是在图上，我们的决策引擎都会将用户所学的知识所定下的决策、用户的习惯进行融会贯通，让用户不仅掌握表层知识，更在无意识的过程中掌握底层的思考框架，成为一个终身学习者。 No matter what space the user is in — in a conversation, in notes, or on a diagram — our decision engine weaves together the decisions their knowledge implies and their habits, so users master not only surface knowledge but, unconsciously, the underlying framework of thinking, becoming lifelong learners.

驰骋千里，世界尽在手中。这里是 Lemma。 Ride a thousand miles and the world is in your hands. This is Lemma.

Oasis Company

github.com/Oasis-Company

12 / 融资需求12 / FUNDING

天使轮融资 Angel round

寻求志同道合的天使投资人，共同打造下一代智能学习平台。 Seeking like-minded angel investors to build the next-generation intelligent learning platform.

¥350万

融资金额AMOUNT

7%

出让股权EQUITY

¥5,000万

投后估值POST-MONEY

8-10 月M

Pre-A 计划PRE-A TIMELINE

股权结构（融资后）Equity Structure (Post-round)

创始人Founders

80%

天使投资人Angel

7%

核心合伙人Partners

6.5%

期权池ESOP

6.5%

资金使用计划（12个月）Use of Funds (12 months)

人员薪酬Personnel

¥150万 · 43%

市场推广Marketing

¥100万 · 29%

研发投入R&D

¥50万 · 14%

场地设备Office

¥50万 · 14%

13 / 路线图13 / ROADMAP

从 MVP 到盈亏平衡 From MVP to break-even

清晰的阶段性里程碑，每一步都有可验证的核心指标。 Clear phased milestones, each with verifiable core metrics.

M0 · 地基≈2-3 周

storage 抽象 → 隐私端点 → pytest。先立信任，再谈智能。Storage abstraction → privacy endpoints → pytest. Trust before intelligence.

价值兑现：隐私可控，用户敢把记忆交给我们。Pays off: privacy stays under user control.

M1 · 记忆内核≈第 1 月

F1/F2/F4/F5/F22 + 伴学接入 + 前端图谱 + 伙伴工坊 v1（第一句话用 AIClient 生成）。F1/F2/F4/F5/F22 + companion + graph + Agent Forge v1.

伴学记得你 + 面板可见 + onboarding 钩子。Companion remembers you + visible dashboard + hook.

M2 · 自适应≈第 2-3 月

F3/F6-F9/F21 + materialization 联动 + 伙伴工坊 v2（绝活+记忆种子）+ LemmaX 考核层框架 + TOEFL 先行。F3/F6-F9/F21 + materialization + Forge v2 + LemmaX framework + TOEFL first.

"会因材施教" + 考试面落地。"Teaches you personally" + exam surface lands.

M3 · 主动智能≈第 3-5 月

F10-F13 + 学习周报 + 伙伴工坊 v3（捏完即活）+ LemmaX 竞赛类。F10-F13 + weekly report + Forge v3 (alive on sculpt) + LemmaX contests.

"会办事的伴学" + 竞赛面。"Agentic companion that gets things done" + contest surface.

M4 · 多模态×生态第 5+ 月

F14-F20 按需：视频理解、图形白板、中文语义检索、多端触达、技能开放。F14-F20 on demand: video, whiteboard, CJK search, multi-channel, open skills.

多端、多模态、生态。Multi-channel, multimodal, ecosystem.

功能全景 · F1–F23FEATURE PANORAMA · F1–F23

T1 终身学习记忆T1 LIFELONG MEMORY

F1记得你Remembers youM1 F2卡点归因Bottleneck diagnosisM1 F3教学规则自进化Self-evolving rulesM2 F4间隔复习Spaced reviewM1 F5知识图谱Knowledge graphM1

T2 自适应学习T2 ADAPTIVE

F6难度自适应Difficulty adaptM2 F7个性化讲解Personalized explainM2 F8学习路径生成Path generationM2 F9动态课程生成Dynamic courseM2

T3 主动智能T3 AGENTIC

F10伴学代理Companion agentM3 F11出题/测验代理Quiz agentM3 F12自主研究代理Research agentM3 F13错题管家Mistake managerM3

T4 多模态×沉浸T4 MULTIMODAL

F14视频理解增强Video understandingM4 F15图形/白板交互Graph / whiteboardM4 F16中文语义检索CJK semantic searchM4 F17语音交互Voice (later)远景

T5 多端触达×生态T5 REACH × ECOSYSTEM

F18多端提醒Multi-channel nudgeM4 F19学习环境生态Ecosystem (later)远景 F20技能开放Open skills (later)远景

T6 认知洞察T6 COGNITIVE INSIGHT

F21学习周报Weekly reportM2 F22掌握度雷达Mastery radarM1 F23效率归因Efficiency insightM3

让每个人都拥有完美助教 A perfect assistant for everyone

感谢您的时间。如果您对 Lemma AI 感兴趣，欢迎与我们联系。 Thank you for your time. If you're interested in Lemma AI, we'd love to hear from you.

moelinkcloud@gmail.com 返回顶部Back to top

          微信 / WeChat: Max53812659
          WeChat: Max53812659
        

别的 AI 教你知识，我们训练怎么教你 Others teach you knowledge.We train how to teach you

为学习场景而生的AI-Native 体验 An AI-Native experiencebuilt for learning

我们不生成内容，我们赋能内容 We don't generate content,we empower it