📡 Idea Radar 每天 8:00 / 18:00 推送 · 18:00 生成精选
reddit · SideProject · 68 分 ← 返回列表

I built an offline flashcard generator that checks its own cards against your source text

I made cardsmith because making flashcards by hand from a reading assignment is slow enough that I usually skip it and just reread, which is a worse way to study. Existing AI flashcard tools need an account and a subscription, or they generate cards with no way to check them against the source before you start memorizing something wrong. cardsmith runs entirely offline. Drop in a PDF, a slide deck, or a text file, and a local LLM through Ollama builds a spaced-repetition deck. Every card ca

五维评分

付费意愿
65
需求真实度
60
个人可落地
90
竞争空间
50
新鲜度
75

AI 分析

痛点:学生手动制作闪卡太慢,现有 AI 闪卡工具需订阅/账号且无法核对来源,容易背到错误内容。

目标用户:需要从 PDF、课件、文本高效复习的学生、考试备考者及隐私敏感的学习者。

证据:Existing AI flashcard tools need an account and a subscription, or they generate cards with no way to check them against the source before you start memorizing something wrong.

变现:开源核心 + 付费桌面版/一键安装包、批量处理、高级 Anki 导出、云同步或学校/团队授权;也可做隐私优先的一次性买断。

MVP:用 Ollama + 本地模型 + PDF 解析,实现上传文档→生成卡片→逐卡引用校验→SM-2 浏览器复习→Anki 导出;两周内先支持 PDF/txt 和单一学科模板。

竞品:Anki、Quizlet、RemNote、Gizmo、Knowt

摘要:作者发布 cardsmith,一个完全离线、通过 Ollama 本地 LLM 从 PDF/幻灯片/文本生成 SM-2 闪卡的 SideProject,并会校验每张卡引用是否与原文一致。现有 AI 闪卡工具多需账号和订阅,且无法核对来源;它在 3,248 词章节生成 152 张卡,40 张人工评分准确率 92.5%。