Arivu's New Direction: A Self-Hosted Second Brain
Arivu has grown from an AI bookmark manager into a self-hosted workspace for web research, notes, research objects, document imports, review, and cited answers.
Arivu started with a simple frustration: saved links disappear.
You find a useful article, save it, and trust that your future self will return. Most of the time, that never happens. The link lands in a folder, a browser profile, a read-later queue, or a bookmark manager. It is technically preserved, but functionally gone.
The first version of Arivu attacked that problem as an AI bookmarking problem. Summaries, tags, semantic search, and resurfacing were the center of the product. That direction still matters. But it was not enough.
The hard part of knowledge work is not only saving and finding. It is deciding what a piece of information should become.
Does this link need action? Does it belong with a note? Should it come back next week? Is it useful today, not useful anymore, or never worth resurfacing again? Does it support a question you are asking now? Can you take it with you if the tool stops fitting?
That is the new direction for Arivu: a self-hosted second brain for web research, notes, research objects, tasks, reminders, review, and cited answers.
What changed
Arivu is no longer just a place where saved links land. It is organized around a working loop:
- Today gives you the daily cockpit.
- Capture saves URLs, documents, transcripts, quick notes, tags, quotes, and standalone notes.
- Inbox lets you decide what each item is.
- Focus gathers open loops from tasks and reminders.
- Review brings back older, due, high-priority, or still-actionable material.
That loop is the product now.
Search, summaries, tags, and the knowledge graph still matter. They sit inside a larger workflow where saved knowledge moves through decisions, actions, and review instead of sitting in a beautiful archive.
Today is the home base
The default signed-in route is now Today. It collects the signals you need at the start of a work session: your dated daily note, Inbox items, open loops, Review items, recent notes, and memory-jogger prompts.
That changes the emotional shape of the app. You do not have to ask, “Where should I start?” The app starts with the day.
Use the daily note for plans, decisions, loose thoughts, and end-of-day wrap-up. It is intentionally separate from the permanent note list. Some thoughts are operating context. Others deserve long-term treatment. Arivu gives both a place.
Notes are first-class
The old mental model was bookmark-first. The new model is item-first.
A saved web page can carry notes, tasks, reminders, annotations, links, and review state. A standalone note can do the same. Notes are not comments attached to bookmarks. They have their own workspace, their own links, their own backlinks, and their own open loops.
That matters because real research rarely starts and ends with a URL. Sometimes you save a source. Sometimes you capture a thought. Sometimes you write a reminder to revisit a source before a meeting. Sometimes you link a note to a bookmark because the connection is yours, not something an algorithm should infer.
Research has named objects
Some knowledge is bigger than a note but smaller than a full project management system.
Arivu now has lightweight research objects for projects, people, books, meetings, decisions, and research threads. An object can keep a title, description, small structured fields, and source context. A meeting can connect to the notes and decisions it produced. A research thread can collect the sources and notes that keep returning to the same question.
This matters because tags alone are too flat. A tag says “launch.” A project object can say what launch means, which decisions belong to it, which meetings touched it, and how the topic evolved over time.
Documents belong in the same workspace
Web pages are not the only research input. Arivu can now import documents and transcript-style material as searchable notes: EPUB, PDF text, plain text, Markdown, HTML, transcript text, pasted OCR, and optional provider-backed image OCR.
Those imports do not become a separate media library. They join the same Inbox, search, Review, export, and backup workflows as normal notes.
AI is optional, not the foundation
Arivu still benefits from AI providers when you configure them. Cited answers, richer summaries, embeddings, and assistant suggestions all get better with a provider such as Gemini.
But the app is designed to remain useful without an AI key. It can still save pages, write notes, track tasks and reminders, run Inbox and Focus, search stored content, import and export data, and produce deterministic local structure.
That is important for self-hosting. AI should be an upgrade, not a dependency that decides whether your second brain works.
The assistant proposes, you approve
Most AI assistants ask for too much trust. Arivu’s assistant is deliberately governed.
It can read context from your Inbox, Review queue, search results, or a specific item. It can draft bounded actions such as changing an item’s stage, creating a link, adding a task, or setting a reminder. But it does not run those actions directly.
Drafts become reviewable proposals. You approve, reject, or ignore them. Nothing changes until you say yes.
That is the right shape for a tool that holds your personal research. The assistant should make decisions easier, not take control of your archive.
Ownership is part of the product
Arivu now runs as one Go application with embedded browser UI assets and SQLite persistence. That is not just an implementation detail. It is a product choice.
Self-hosted software should be boring to operate. The deployed frontend should not need a separate Node dependency tree. The database should be understandable. Backups should be practical. Provider integrations should be optional. Exports should be real.
Arivu supports JSON backups, CSV, browser HTML, Markdown, and Obsidian-ready ZIP exports. Imports can bring in common bookmark formats, feed lists, URL lists, documents, transcripts, calendar events, and Arivu backups. Your knowledge should not become trapped because the tool was useful for a while.
What this means for older Arivu posts
Some older posts on this site describe Arivu as an AI bookmark manager or capture layer. That was true for an earlier direction. The current product is broader.
We are keeping those posts for historical context where useful, but the canonical framing is now this:
Arivu is a self-hosted second brain for web research, notes, research objects, tasks, reminders, review, and cited answers.
If you want the practical workflow, read How to Build a Second Brain with AI Bookmarking next. If you want the daily operating model, read How to Run a Daily Second-Brain Workflow.
The old problem remains: saved links disappear. The new answer is stronger: give every saved thing a path through capture, decision, action, and review.
That is where Arivu is going.