MindManager, evolution regarding LLMs

1. on filelinker - this is only a modified copy-paste to me

2. it breaks - https://community.mindmanager.com/topic/2780-filelinker-help-some-but-risk-of-data-loss

3. others esp Robert made some advancements, https://github.com/robertzaufall/mindm/; even i myself made some. however, if the giants like @Nick didn't join, it's too heavy for others to pick up. i asked openclaw(indeep chinese's autoclaw)+GLM to re-write mindm using native COM APIs, which will be much faster (on windows and windows only) than mindm. however, it's working thru REST APIs. further re-writing it into an addon will be even much more faster. but it's too heavy for a layman like me to do so. installing vs studio (not vs code) and compile those addon is nightmare. my bet later is if github could provide the tools for me to compile directly, instead i have to install on my laptop.

4. another problem i worry is monetizatoin.

if i were @nick, i could prepare many things using LLM to charge, while providing professional grade products.

a. tons of ppl want dark background in topic note, as well as markdown viewer. my latest post showed that we could run a small python server (now outside mindmanager, but indeep be integrated if it'as an ADDON) and then we could have markdown reader, and topic note dark theme reader.

b. the native COM APIs that autoclaw opened me to re-write (alike mindm), could be made into an addon, which solve OP's problem. currently as a layman i am using an external small python web server way, which is slower and just make ppl feel it's less professional and polished.

btw, .mmap could be turn into .json by mindm or my CLI, that if someone want, he/she could build it in 1-2 days and polish in weeks. but vibe coding is just not that professional.

tell you what, with mindm (or my CLI) i could now export the .mmap easily into .json, that include almost everything one want to migrate to other mindmap programs. it's just at at this moment, mindmanager is still the most polished mind map program so i stay.

important features that other mindmap programs sucks

1. mm is 1 file 1 map, this is important; xm1nd, the most aggressive competitor, is 1 file many maps.

2. 1 file's 1 map's 1 topic can hyperlink to another file's map's topic, using mj-map://; freep1ane can do so, but many others cant.

btw, file linker may be the answer for web chatbot, but what need-ed now is agentic ability. that's the LLM do things on behave of you. like in openclaw.

Thank you for your generous words. I am working on an AI-related application of MindManager, but it does not involve integration with LLMs or agents. Personally I am not convinced of the value of inserting AI-generated content into maps. I see mind maps as an antidote to AI-generated content because rewriting and restructuring is where thinking happens. What would be really valuable is a summarisation of the contents of a mind map, especially a badly organised one. This is a non-trivial issue as the structure of a map is what gives it meaning. AI needs to discern that structure, recognise anomalies, and make an educated guess about what the author meant. Assuming that a map is a logical outline is not enough. If I could throw hundreds of maps into a database then ask AI to find stuff that is related to or similar to a given concept, that would be transformational.

Thanks for the reply.

"If I could throw hundreds of maps into a database then ask AI to find stuff that is related to or similar to a given concept, that would be transformational."

1. This is exactly where I'm heading. I want MindManager to work like Obsidian — and it's happening. I've written several add-ons that export topic text into .mmap.md files, placed inside a folder for boundaries (like Obsidian vaults, not random files scattered on SSD). Then LLMs can link them together automatically. This is the agentic branch of LLM — not the "knowledge" branch. Think of them as cheap labour that works on your behalf.
2. Separately, I've vibe-coded some things people here have been asking for:
   Markdown viewer inside the internal browserTopic notes rendered as rich text on a dark background, in that browserEditors (not just viewers) for the above is possibleA PDF reader should also be feasible

The engine (mm_20ultra) is ready. I dont necessary need to monetize it, but i will share the code only if i could get some benefits e.g. if a community will work on it to make it polished and workable, bug free etc.

it's named "20" for mm2020 as it's the last version that dont need connect to internet. later versions that need phone home monthly is simply spyware in some ppl 's eyes. these software could scrap ur SSD/HDD for a month, and then send them home once per month on that day.

thanks