MindMax v1.1 covers three builds: 1.1.0 and 1.1.1 on 12 October and 1.1.2 a day later. Most of it is about two things: opening more of the files people actually have, and letting AI work on them without taking the file out of your hands.

What is in it​

• PowerPoint decks open as slides. A .pptx opens as the deck itself, and the AI can read it slide by slide, answer about one slide or a region you box on it, and leave notes page by page. Lecture slides and pitch decks were the most common files people could see in the folder but not use.
• Translate a deck and keep its layout. A translated .pptx comes back as another .pptx with the same themes and positions. Only the words change, and slides that may overflow are flagged for you to check.
• Kindle books. .mobi and .azw3 files open as books, with their table of contents, and work like any EPUB for reading, asking and translating.
• AI edits inside Word files. Ask for a revision of a .docx and the accepted edits land in that file as Word tracked changes, so whoever opens it in Word sees them the usual way. A PDF cannot take edits, so it gets an editable Markdown copy beside it. An AI review of someone else's draft can also go out as Word comments in a copy of their file. Most documents worth revising arrive as Word, not Markdown.
• Quick notes and voice. A floating capsule takes a thought typed or spoken, from anywhere in the app, and files it into the note you have open or today's daily note, as a change you accept. The cost of writing something down has to be close to zero, or it does not get written.
• Images in the app and on the canvas. Images open in a viewer that can describe them and answer about a region, and they can sit on a canvas as cards. Paste text or an image straight onto a canvas.
• Links panels for notes. Each note shows what links to it and what it links out to, including mentions that are not linked yet, with a preview on hover. A folder of notes becomes a web you can walk.
• More you will notice day to day. Several workspaces you switch between, a native application menu, emoji in the editor, card positions you can pin on a canvas, a redesigned comments panel, and model settings that read the available models from your provider instead of a fixed list.
• First-run guides in your language. A first-time folder now comes with guides in Chinese when the app is in Chinese.
• Microsoft Store. On Windows you can now install MindMax from the Microsoft Store, which signs and updates it for you.

Known rough edges​

• Installers outside the Store are not code-signed yet. On Windows, the .exe shows a SmartScreen warning the first time — choose More info → Run anyway. On macOS, open it once, then allow it under System Settings → Privacy & Security. Unsigned Mac builds cannot install their own updates; they tell you when a new version is out. The installation guide walks through both.
• Translated decks can run long. Some languages need more room than the original text box. Check the slides flagged as possibly overflowing before you present.
• Plans changed. What each plan includes, and how weekly credits work, were revised in 1.1.1. The pricing page has the current terms.

Where to start​

New here: open your first folder. Editing a Word file: edit Word documents with AI. Already using it on Windows or from the Store: the update will offer itself — accept it and restart.

用 AI 整理笔记最安全的方式，是让它改笔记「怎么写」，永远不改它「说了什么」。MindMax 的快记就是这样工作的：每一个错别字、听错的字和破碎的句子都会被修正，想法被归到对的标题下，而你没说过的东西一样都不添——不添事实，不添结论，也不把话说得比你更满。

这听起来是个小区别。它却是「一篇下个月还能信的笔记」和「一篇得重新核对的笔记」之间的全部差别。

想让 AI 以修订模式修改一份 Word 文档，AI 就得修改 .docx 本身，而不是交回一段改写过的文字。在 MindMax 里，你 @ 这个文件，把要改什么说一遍，分组审阅修改，你接受的那些会以 Word 修订的形式写进同一个文件——样式、编号和页眉原封不动。

这听起来是个小细节。它却是「一份能直接发给客户的修改」和「一个下午都花在把一份原本好好的文档重新排好」之间的全部差别。

想用 AI 审阅一份 Word 文档、交回真正的批注，AI 就得把每条批注锚定在它针对的那几个字上，而决定发出什么的，得是你。在 MindMax 里，你说明这次评审为了什么，把你认同的起草批注用你自己的话采纳下来，再把它们以普通 Word 批注的形式导出到 .docx 的一份副本里。

评审的意义不在于看法本身。而在于一个作者能据以行动的看法，落在它针对的那句话上，在作者本来就在用的那个文件里。

想就一整本书向 AI 问一个真正的问题——论证在哪里转向、作者收回了什么、第九章是否和第二章矛盾——AI 就得像你那样读这本书：看目录，判断问题需要哪几章，把它们完整读完，第一章没解决就回头再读一章。把整本书贴进去，和搜索切碎的片段，恰恰都在这些问题上失败。

每个 AI 阅读工具都得决定：当材料比模型能好好处理的更长时，该怎么办。这个决定在回答里是看不见的，它却决定了这个回答能不能信。

MindMax 本身可以免费使用：阅读、写作、搜索，以及把文件留在你自己的文件夹里，都不花钱；你唯一要付费的是 AI 的活，要么通过订阅，要么用你自己的 API key。打开一份 PDF、高亮、搜索你的文件夹、编辑一篇笔记、导出为 Word，都从不消耗额度。提问、讲解、批注、翻译、识别扫描件或者转写你的语音，是 AI 的活，会消耗。

这条线是刻意划的，这篇文章解释它划在哪里、为什么——因为在大多数软件里，「免费版」的意思是「一个第二天就用完的演示」，而我们指的不是这个。

决策矩阵是一张表：选项排在侧边，标准排在顶部；你给每个选项在每条标准上打分，可以给标准加权，然后比较。它的价值与其说在最后的总分，不如说在这份纪律：每个选项都用同一把尺子衡量，而那条你一直悄悄忽略的标准，变成了一列你必须填的格子。

「思维模型实战」系列的最后一篇：一个关于经典方法、以及怎样把它们用在你自己材料上的短系列。

核对一句 AI 引文是否真的出现在 PDF 里，听起来就像按一下 Ctrl+F。实际上，PDF 里的文字并不是你看到的文字：单词在行末被连字符拆开，段落被分页和页眉切断，像「fi」这样的字母组合被存成一个符号。一个无视这些的引文检查，要么拒绝真实的引文，要么放过改写过的话。把它做对，就是大部分的工作量。

我们说过，MindMax 里的每一句引文都会对照原文核对。这篇文章要讲的是，为什么这句话背后藏着的比它说出来的多——以及你应该对任何作出同样声称的工具抱有什么期待。

六顶思考帽是一种从六个角度依次看同一个问题的方法：白帽看事实，红帽看感受和直觉，黑帽看风险，黄帽看价值，绿帽看新点子，蓝帽管理过程、决定下一步。它的要点在于分开——不在同一句话里让乐观和谨慎互相争论。

「思维模型实战」系列之一：一个关于经典方法、以及怎样把它们用在你自己材料上的短系列。

想用 AI 写学位论文的一章而不丢掉论证：在起草任何正文之前，自己定好论点和大纲；依据你自己的资料写，而不是模型的记忆；随时记下草稿已经论证过、定义过、承诺过什么；每一处 AI 修改都以带标记的修订来审阅。另外：动手之前先查清你所在院校关于使用 AI 的规定，而不是写完再查。

常见的做法——让对话助手一次写一节——写出来的每一节都还行，整章却立不住。