When meetings and notes become knowledge

Remote work produces a lot of meetings, and meetings hold an astonishing amount of knowledge — decisions, the reasons behind them, the half-commitments, the tone. I used to have two options: listen, or take notes. Doing both well was never really possible. Take notes and you stop thinking along; think along and you lose half the detail.

Fireflies takes over that part now. It records the meeting, produces a transcript, and then summarizes the conversation. The immediate effect is unspectacular and still large: I can concentrate on the conversation again instead of on my notes field. The note-taker role I had been quietly playing on the side is gone.

These AI meeting assistants can do a surprising amount by now — tell speakers apart, pull action items out of the discussion, separate the substance from the side conversation. You should still know where their limits are. An automatic summary smooths things over; it sometimes loses exactly the one offhand remark that turns out later to matter most. And because conversations are being recorded here, data protection is not a footnote: whoever records carries the responsibility that everyone in the room knows about it and is fine with it.

For me, though, the transcript was never the goal. It is only the start. A transcript that sits in some tool and is never opened again is as worthless as the meeting no one remembers. The value only appears in the next step — when that information is processed further and lands somewhere it will stay. And that is where the next tool begins.

Obsidian is my digital brain. That sounds bigger than it is, and it still fits best. What lands there is not just documentation but knowledge — and that is a difference. Documentation describes what is. Knowledge connects why something is the way it is and how it relates to everything else.

A large part of it flows in automatically: Fireflies feeds the meeting summaries straight in. To that I add decisions, project notes, product knowledge, ideas, insights. Piece by piece a knowledge base forms that gets a little denser every week. For complex products this is enormously helpful. Anyone who has tried, six months later, to reconstruct why a feature was built exactly this way and not another knows the problem: the knowledge was there, and then it evaporated. It sat in a chat, in someone’s head, in a meeting no one remembers.

Product people have a term for this that fits well: the second brain. Personal knowledge management, a memory outside your own head. Tools like Obsidian are built for it — not as a linear folder tree but as a web. Through bidirectional links a knowledge graph emerges, in which notes point at each other and connections surface that you never had in mind while writing. Information doesn’t have to be searched for over and over; it sits there, linked.

The real lever comes only now. You can search this knowledge base with AI — not by keyword, but by meaning. An archive turns into a conversation partner that knows everything I have ever written down. That is why, for me, this is not a note-taking app but a genuine store of knowledge. And it has a property that has become rare: the longer I work on it, the more valuable it gets. Most tools are at their best on day one and become a burden afterward. This one is the other way around. Knowledge kept cleanly over years is one of the few real competitive advantages you cannot simply buy in a hurry.

Over time this becomes a real store of knowledge. But a store is only worth as much as whatever puts it to use. That is where the actual engine comes in — the next part.