What your second brain keeps, and what you only said out loud

An AI second brain is a notes system you keep outside your head, with software that helps you find things in it. It holds what somebody wrote down. What was said out loud — a decision on a call, a conversation across the table, a thought you dictated to a model — gets there only if somebody types it up afterwards. Dalm covers that part. It records calls and conversations in the room and takes dictation on your Mac; the speech is recognised on your Mac and the audio is never sent anywhere, while the text is synced to our servers. Your notes stay where they are, and what you need comes out as Markdown.

What a notes system already does well

If you keep one — a notes app, a folder of text files, pages that link to each other — it works, and nothing on this page argues otherwise. You chose what was worth keeping and wrote it in your own words, and that is exactly why you can find it again: you search for words you used yourself. It is yours, in a format you picked, and it outlives any one tool you feed it from.

Where it stops is not a flaw in how you built it. A notes system holds what somebody typed. For what to do with the notes you already have, see why filed notes get lost, and what to mark instead.

What never reaches it

A decision made on a call. A number a client said across the table. A phone call on speaker. An idea you talked through with a model instead of typing it. All of it reaches a notes system the same way: someone types it up afterwards — and that is the step that slips when the next meeting starts.

Dalm records a call from your Mac without joining it, a meeting in the room from the microphone, and a phone call on speaker the same way as a meeting in the room. If you are choosing between tools that record meetings, the kinds of AI note taker are laid out side by side.

Where Dalm fits: feeding the system you keep

Dalm does not ask you to move your notes. It works on the part that never reached them.

During a meeting you can still write your own short lines. They are the frame of the summary: it is built around what you wrote rather than retelling the recording. Under each line, Dalm shows what was being said just before you wrote it — taken straight from the transcript, not written by a model.

Everything you say lands in one archive you can search: meetings, phone calls, conversations in the room, and what you dictate to a model. You search it by the words that were said; asking one question across all your records is in development. How a product manager uses it from one sprint to the next is on the page for product managers. If you already keep a research notebook that answers from files you upload, see how a research notebook and Dalm compare.

Getting it into your own system is ordinary. It comes out one document at a time — there is no export of everything at once. And dictation types into whatever field the cursor is in, the page of your notes app included: dictation that types into any app.

What the AI part does today, and what it does not

Summaries. After a meeting, a language model writes the summary — decisions, action items, open questions — and leaves a field empty rather than invent a deadline or an owner nobody named. Every item carries the words it came from. Every quote shows you where it was said — open it and the transcript opens at that place.

Questions. You can ask about one meeting, and if the answer is not in it, you are told so. Asking one question across all your records is in development. What already works across meetings is two lists, Tasks and Decisions, gathered from your meetings.

Search. One field searches meetings and dictations at once: transcripts, titles, descriptions and the documents built from them. It is a full-text search: it matches the words you type. The short notes you type during a meeting are not part of the search.

What goes where. Speech is recognised on your Mac and the audio is never sent anywhere. The audio is deleted once the text exists. The text — transcripts, your notes, documents and dictations — is synced to our servers, and when a summary is written or a question answered, that text goes to a language model; the audio does not. Which steps run on your Mac and which on a server are laid out step by step.

Free, Pro or Team: where to start

Dalm has a Free plan. It is a plan rather than a trial, and it does not run out. You need a Mac with Apple Silicon · macOS 15.0 or later and an account — the app does not open without one — and while the app is not public, the way in is the waitlist. What the paid plans add, and what they cost, is on the pricing page. If several people need their meetings in one shared space, working as a team explains how that is set up.