An AI second brain that keeps your notes up to date

Building an AI second brain usually goes one of two ways. You install an AI plugin for Obsidian and ask questions about your vault. Or you point Claude at a folder of markdown and write instructions for it to tidy and connect your notes. Either way the notes are still yours to keep up: the AI answers when asked, and the filing, updating and remembering stay with you.

A second brain that actually takes work off your hands does three more things. It adds to your notes without being asked. It runs on a schedule, so the weekly summary is there on Friday. And it shows you every change before it lands, so you can trust it with the folder.

This is how notes work in Aplite.

Plain markdown, one tree

Every workspace is a folder of markdown files: your notes, your lists, the plans for the week — and next to them the agents’ own instructions and memory. Headings, checklists, tables and [[wikilinks]] work the way you expect, and pasting markdown in keeps it markdown. Nothing sits in a format only the app can read.

The agent reads the same files you do. Ask it what is left on a project and it answers from the note, not from a guess.

It writes, you accept

When the agent wants to change something — two lines in today’s plan, a new row in a list, a whole new note — it proposes the change. The new lines appear right in the note on a green rail with Accept and Reject next to them. You can accept, reject, or edit the proposal first. Nothing is applied until you say so, and every proposal also waits in Approvals in the header.

When a task is ambiguous, it asks one question back instead of guessing: plan the weekend outdoors or around a concert?

It remembers in the open

What the agent learns about you goes into its own memory notes — dated files in the same tree, which you can open, read and correct. “Film night is Friday” is a line in a file, not a hidden setting. If it remembered something wrong, you delete the line.

What to ask it

Each of these is one sentence. The agent creates the task, the schedule and the note it files into.

Every Friday, check arXiv for new papers on my topics and file two-line summaries in my notes.

Copy

The arXiv keyword watch keeps a reading list you never have to update yourself.

Each evening, transcribe the new audio in my workspace and file it as clean notes.

Copy

Voice memo transcripts: record on the go, read tidy notes in the evening.

Every Friday afternoon, turn this week’s notes and numbers into a short deck.

Copy

A weekly status deck built from what you already wrote down.

Every morning at eight, send me my day on Telegram — calendar, weather, and what’s waiting in my notes.

Copy

The morning briefing reads your notes so the reminder comes to you — see an AI assistant in Telegram.

Bringing your notes in

Create notes in the workspace, or paste them in from wherever they live now — a note pasted as markdown arrives with its headings, lists and links intact. Start with the notes you actually work from: the current projects, the running lists. The agent gets useful as soon as there is something real to read.

Your notes, your models

Each workspace runs in its own container, on a server set up for you in about a minute — no vault sync to configure, no API keys to wire into a plugin. The agent runs on any model from the catalog, and you see what each answer cost. Once you trust it with a folder — a reading list, a journal — a standing rule lets it write there without asking; everything else still comes as a proposal.