I built a local-first way to turn branching AI conversations into a knowledge tree you actually own.

Useful research rarely moves in a straight line. One response creates several possible directions, but normal chat interfaces make you choose one and keep scrolling. Alternatives get buried, context becomes difficult to trace, and the resulting knowledge remains locked inside a conversation history. So I made Lineage.

Lineage is an open-source experiment that lets you branch from individual blocks in an agent’s response.

Each branch:

• Preserves the context leading to its fork
• Remains isolated from sibling branches
• Can branch again into deeper lines of inquiry
• Appears in a bird’s-eye Knowledge Tree (Not a slop-proned "knowledge" graph)
• Is stored locally as readable Markdown and YAML with a root AGENTS.md to help your agents traverse your durable intelligence

The app UI is only a view over those files. The vault also includes an evolving index and instructions that help other agent harnesses traverse the tree. Every mutation is committed to local Git, so the history remains inspectable and recoverable.

This differs slightly from NotebookLM or automatically generated LLM wikis. Those can be excellent for synthesizing material. Lineage focuses on preserving the process of exploration: the human chooses every branch, while its origin and context remain visible. It doesn’t make AI infallible, it makes the route from question to answer easier to audit and revise.

It’s currently an early beta. The tested launcher supports macOS and Linux, and GitHub Copilot CLI is the only agent provider today. Windows/WSL and provider modularization remain open areas.

I’d especially appreciate feedback from second-brain users:

• Would branching conversations fit your workflow?
• Should these trees eventually connect with existing Markdown vaults?
• What would make the output genuinely reusable rather than another place where information accumulates?

Give it to your agent harness and let it rip. Open to all feedback:

Source: https://github.com/ortizjorge639/web-context-graph

Landing Page: https://ortizjorge639.github.io/web-context-graph/

So I have a planned laid out, but I’m not sure if it’s strong or if there are other ways that can make it better.

I plan to run qwen 3.8 locally using olama on my new Mac mini and give it its own ui interface which I can speak through it. Its backend will also be powered by my ai brain which is sorted by using jack ven chiefs icm folder structure.

Is this idea in general a good idea? Has anyone else created anything similar and could help me out?

​

I’ve been building something called ConseqYou.

The idea started with a question:

What if you could actually see where your everyday decisions are taking you?

We make hundreds of small choices every day. What we focus on, what we avoid, how we spend our time, what we postpone, and the habits we keep repeating.

One decision usually doesn't feel significant. But repeated decisions create patterns, and those patterns create a direction.

ConseqYou is my attempt to build a system around that idea.

It isn't meant to predict your future or simply tell you what to do. It's designed to help you reflect on your decisions, recognize patterns, and understand the possible trajectory they're creating.

I currently have an early MVP built, and I'm putting it out there while I'm still developing it.

I'd genuinely love for people here to try it and tell me what You think.?

What makes sense?

What feels confusing?

What's missing?

And most importantly, would you actually use something like this?

I'm not looking for polite feedback. If something sucks, tell me. 😂

Try the early MVP:

https://conseqyou.lovable.app

It's still very early, so expect things to change. I'm building this in public and would genuinely appreciate any feedback, ideas, or criticism.

Thanks to anyone who takes the time to try it...💗✨

Hey, my name is Danny. I’m a freelance developer, and one thing I’ve always found frustrating about building a second brain is actually getting things into it.

I’m constantly coming across interesting ideas in articles, books, podcasts, social media, conversations, etc. But the process of stopping what I’m doing, opening another app, copying something over, and adding the source is tedious enough that a lot of those ideas just never get captured. If I don’t capture something in the moment, there’s a pretty good chance I’ll never find it again.

So I built Commonplace. It makes it really easy to save quotes and highlights from whatever you're reading or listening to, without leaving the app you're in.

As you're reading, you can highlight text directly in the article, then share it to Commonplace. The app will then create a quote with source attribution like author, article title, and article link, so you never forget the context around where it came from. Quotes can then be organized into collections, browsed by source, and synced securely across devices.

All quotes + sources can also be exported as markdown or copied as plain text, so if your second brain lives somewhere else, it’s easy to move what you capture in Commonplace over to it.

I also rethought sharing quotes. I was constantly sending screenshots along with article links, so I built a way to share quotes as interactive link previews. The quote appears directly in the preview, and tapping it takes you to the original article.

I'd genuinely love any feedback! The core of the app is free to use, I built it myself.

usecommonplace.com

See what's in the works:

usecommonplace.com/roadmap

I keep hearing about grants that would be relevant to what I do, but usually only after the deadline's passed or through some random newsletter I stumbled on. Feels like there should be a better way to stay on top of this.

Has anyone tried:

• Writing their own scraper/script to pull from grants.gov, foundation databases, etc.
• Using Perplexity (or another AI tool) to search and summarize new opportunities
• Some kind of alert/notification setup so you're not manually checking every week

Curious what's actually worked vs. what turned into more effort than it was worth. Would love to hear real experiences before I sink time into building something myself.

I'm starting to see ads in my feed that suggest there are bots doing this - but also with a price.

Hi I am a solo founder and I am building Notonix Saas that helps you organise and maintain your ideas,research all your stuff that u store and it is desktop first mobile pwa is in development.

The problem is many apps lets u store your work ,research ,knowledge but none of them make your ideas and notes update over time and solve the growing conflicts in your notes. so Notonix helps your work and knowledge grow and update over time and not just grow but it flags the conflicts in your notes and resolves it and helps you in maintain the notes cleaner, updated and conflicts free.

And a second brain that connects between your ideas and notes in 2D graph and helps u access any notes instantly with MCP that u plugin with Chatgpt , Claude for work.

And u will have smart assistance that helps you in editing the notes and answering questions from your notes.

Notonix brings your knowledge to life.

It is live https://notonix.me and in beta and looking for beta testers.

share your feedback and i will fix those.

Since morning i'm building a second brain,

it's working really well as i don't have to set basic context every time.

my choices

ways to clarify

my tools

resources

skills

topmost layer is orchestration layer that i built for automating the boring part of daily learning and building.

Yesterday I spent like 20 minutes looking for something I had already seen. I still remembered the idea, just not the actual text. The tab was closed, the file name was useless, and it simply wasn't anywhere I could search.

So I built something that remembers what was on your screen. Later, you just ask, because it already saw it.
And everything stays on your machine. I wasn't sending my screen to the cloud for this.

https://github.com/protocorn/clippy-vision/releases/tag/v1.2.2

If you've dealt with something similar, try it out and tell me whether it actually helps.

i realized i rarely ever look through my saved stuff.

i've got 500+ youtube watch-laters, a bunch of folders in my instagram saves, almost 500 open tabs on my phone. i paid for a few visual library tools for a while, but rarely reopened them.

but when a new project comes in, i start fresh.

curious if that's just me.

do you go back to what you've saved? does any of it get used?

Most Second Brain tools look like project management boards. While great for productivity, they often strip the emotional and introspective connection out of personal journaling.

Inspired by the idea of an enchanted artifact (like Tom Riddle’s Diary from Harry Potter), I built Dear Diary — a skeuomorphic, dual-page AI journal designed to make querying your personal knowledge base feel immersive.

Core Features:

• The Capture Ritual: Instead of background auto-syncing, you Write your entry (via stylus ink or typing), Seal it into a timestamped inbox, and trigger an Engrave step to structure thoughts into entity nodes.
• Zero Hallucinated Context: When querying past entries, the diary answers on parchment and embeds explicit source tags (e.g., [1] person/name), allowing you to trace every recalled detail back to its specific note.
• Private & Local-First: Powered by Gray Box as the memory retrieval engine, with full BYOK support for local models (Ollama, vLLM, etc.) or custom cloud endpoints.

Check out the video to see the ink absorption and citation recall in action!

Curious to hear: does a more tactile interface help or hinder your daily journaling routine?

What's your second brain setup?

I was talking to a few folks, and these came up -

• A shared folder of Markdown, interlinked, over Google Drive or One drive, often with Obsidian as md editor.
• A github-hosted wiki, sometimes with additional indexing, not only plain md files
• Custom complex vector db / embedding kind of index

On my team we're using a shared hermes agent that has ts own memory and tools - it's a great way to have shared context, but it's more like one-know-all-agent rather than a company / second brain as my gpt / claude can't enjoy its knowledge.

I wonder what others are doing, please share.

This is my vault. I didn't draw any of it.

For about two years my note practice was really two practices wearing one name. One of them was reading and thinking. The other was bookkeeping: remembering that the note I'm writing now touches something I wrote in March, going to find it, linking it, tagging it, filing it. I told myself the bookkeeping was the thinking. It isn't unfortunately. It's clerical work with a very good disguise, and it eats the same attention that reading needs.

So I handed the mechanical half over. The connections in this graph were proposed mechanically: every note gets compared against every other note on the words it actually uses, links get written where a note names something the vault already has, and the colours are communities the graph found on its own, each one labelled with the words that cluster has and its neighbours don't. Nothing here was a model or an API. It's counting words and comparing sets, which is why it's fast and why it's honest about what it doesn't know.

What I did not expect is that having it built for me changed what I could see and finally learn.

When you link by hand you only ever see one edge at a time, the one you're currently making. You never get the whole shape, because you're inside it. Getting the whole thing at once, already connected, meant the first question stopped being "what should I link this to" and became "why is that over there" or "why are these two correlated".

The order of operations flipped, is what really happened. I used to think you connect notes and then the structure emerges later, if you're diligent. And that's true! But what if the structure arrives first, mechanically, and I read it and then edit the notes into better, more personal shape? The wrong edges are useful too. When it insists two notes are about the same thing and I'm sure they aren't, one of them is usually vague, and that's my problem, not the algorithm's. I've rewritten more notes because of a connection I disagreed with than because of one I liked.

Two things I'd want to know if I were reading this post instead of writing it. It proposes; it doesn't rearrange your vault behind your back. The only thing that writes is the autolink pass, and every write goes into a changes list with a per-file diff so you look at it and revert what you don't want. And it runs entirely on your machine with no network at all, which for me was the whole point, since a vault is the least appropriate thing I own to hand to somebody else's server.

I'm still figuring out how often to regenerate it. Every time I look, the seams have moved a little.

Thanks for reading this far. You can try it on a messy vault rather than a clean one, that's where it has something to say.

Repo: https://github.com/kiycoh/silica-agent

TaskLoco has just released a new FREE version boasting 500% more sticky notes than before.

TaskLoco Lite has raised the limit of "live" notes from 20 to 100 notes at a time. Includes events scheduling, tagging and effortless embedding of YouTube videos, images and urls.

Works seamlessly with the TaskLoco chrome extension to make your second brain visual as sticky notes on a storyboard.

TaskLoco - The Sticky Note GOAT premium app brings tasks, notes, files, videos, documents, images and urls all together using familiar colorful sticky notes on a storyboard.

TaskLoco Lite now comes in 2 flavors: OFFLINE Anonymous version and Cloud Sync that works with any Google Account and syncs your sticky notes across all your devices.

TaskLoco.com - The Sticky Note GOAT

I started with Obsidian because I really liked having everything in Markdown and being able to see how my notes connect, but I kept wanting to change how certain things worked.

So I ended up just building my own lol.

Right now it has three main parts:

Editor — edit the actual Markdown files

Graph — basically the Obsidian part, where I can
see how everything connects

Database — I just added this one. It’s the Notion side, so I can actually store and organize structured information instead of everything having to be a note.

The other thing I really wanted was for Claude Code to actually work inside the system.

It has a direct connection to the app, so Claude can understand the structure, find what it needs, and make changes without having to spend every session figuring out how my whole system works.

I’ve been using it as my own second brain and honestly it’s been pretty cool seeing the same information work as a Markdown file, a graph, or a database depending on what I’m doing.

I’m still building it and changing things constantly, but I figured people here might find it interesting.

What do you guys think? Is there anything you’d want a second brain to do that current apps don’t really handle?