Note-taking in LLM environment

Notes

A goal of note-taking is to have notes at the end. Something that will be a result of the meeting, something to easily come back to.

This is what can be easily automated right now.

We can generate a meeting transcript, then feed that to an LLM and have a summary. Many tools automate that whole journey. That is not the main point of note-taking.

I would argue that the goal of note-taking is not the text.

Crystallize your thoughts

When you note something down, you’re making decisions about what’s important enough to record. LLMs can summarize content, but this generated summary won’t force you to take effort, consider important details, create a high level plan on what would the next steps be.

Taking notes also reveals gaps in your understanding. If you struggle to describe a concept in detail to write it down, that’s a signal you need more data.

Different formats of notes

LLMs are, at least for now, quite good at making lists, summaries, turning long texts to small texts or the other way around. Notes can take other forms depending on the needs. If we’re talking about high-level problems, designing the flow of data or figuring out how

Context

Meetings or problems do not all exist on the same level of abstraction. LLMs are fully capable of taking previous notes into account, but not yet to understand what level of abstraction will be most useful. This requires understanding of the problem not in the vacuum but in the context in which it’s being solved.

Timeliness

Most automated tools get you the notes and summaries only after the meeting. Those are valuable as a reminder of what was discussed or for absentees to catch up. During the meeting, by the very nature of generating the summary only after the meeting, cannot help during the meeting.

By taking notes manually, you may find areas of ignorance which may significantly impact the flow of discussion. More meaningful questions to reduce the unknowns or find problems and eliminate them before any time is invested.

Automate the boring stuff

I find it best to take notes:

• When I don’t have a clear idea of the problem
• On complex problems with multiple solutions
• When there is not much time, as a way to keep on-track

Everything else? Keep it in cache, check LLM Summary straight after the meeting. If anything of importance is missing, I add some notes, ideally sharing with others if that makes sense, so everybody can benefit.

Try taking notes for a week and share them with the attendees. Most people will appreciate them and you will get the additional benefits too!