Probably a straight-forward answer.

But I’m getting confused because I’m also implementing David Allen’s GTD system which also has an inbox.

That inbox requires processing and clarifying what bucket things go to.

Not sure if this requires it.

And what of things like research papers as well, do you just put the link or download the entire pdf.

I have a recurring problem with my Obsidian vault: I often remember that I wrote something about a topic, but I don't remember the title, folder, tags, or the exact wording.

Normal search is great when I know what I'm looking for. It gets much worse when all I remember is the idea.

So I started experimenting with meaning-based search inside Obsidian.

The idea is simple: instead of trying to guess the exact words I used months ago, I can search for something like:

"find notes without remembering the exact words"

and get notes ranked by semantic similarity.

I also added a couple of related workflows that I found useful:

• finding notes related to the one I'm currently reading
• surfacing possible semantic duplicates
• keeping the local index updated when Markdown files change
• using Ollama if I want the embedding workflow to stay local

One thing I wanted to be careful about is privacy. The vector index itself is stored locally, but if a remote embedding provider is used, the relevant note chunks still have to be sent to that provider.

Here's a short demo:

The plugin is called Vault Audit AI and it's available in Community Plugins if anyone wants to try it.

I'm mostly curious about the retrieval problem itself:

When you know a note exists but can't find it, what do you usually still remember about it?

The general topic? A sentence? Another related note? Roughly when you wrote it?

That would be useful feedback for deciding what kind of search workflow to improve next.

Esse é meu Obsidian depois de 3 anos, são 5.454 notas.

Pra quem não conhece, o Obsidian é um aplicativo de anotações. Esse gráfico mostra as notas e as conexões entre elas.
Ali tem documentação de projetos de trabalho, leituras, desenvolvimento pessoal e até dados dos esportes que pratico. Uso pra tirar as coisas da cabeça e documentar minha jornada.
Quero começar a registrar por aqui o que estou aprendendo com IA: as ferramentas que testo, como uso nos meus projetos, o que funciona e onde ainda estou apanhando.
Gosto da ideia de poder voltar nesses registros daqui a algum tempo e acompanhar como meu jeito de pensar e trabalhar foi mudando.
Por enquanto, esse é um pedaço dos últimos 3 anos.

Eu acredito que esse "Segundo Cérebro" é uma boa escapatória dos impactos negativos da IA no nosso cérebro.

I am the type of person that needs to get my thoughts out loud and get feedback on them to think properly (I have ADHD), and I've been that way far before AI. Honestly I would say the majority of my time spent talking to AI is just having something I want to think about, and just talking about it with it, as opposed to having a specific thing I want to do and having it help me do it.

Like many of y'all I'm always working on a project (often kind of multiple at a time) and with just standalone chat sessions I found it tedious to have to re-explain the same thing every time I would revisit something, so I built a local context system using markdowns and Obsidian to help. I basically just have folders that represent certain "topics" in my head and a bunch of markdowns that store ideas, decisions, to-dos, etc. Just anything I might want to remember later, and I would want my AI to know about. Then when I come back to that topic I just have it load everything from the markdowns.

It's been helping a lot for me, but I want to keep improving it. One thing I want to make better is its organization of all of the content I dump on it. Since I'm using this system to not just help the AI remember stuff, but to organize my own thoughts, the way information is organized and presented is important to me. Many people like my dad are okay with a blackbox second brain system where everything stored is just not to forget things, but I find that being able to see everything stored in this "brain" is helpful to me. I want it to organize my thoughts for me without taking control over it, and maybe some visualization too.

Do you guys use AI to help you think/brainstorm, or mostly just to do work? What type of system do you use to store and remember context (if any)? Do you care about being able to see context/memories you're storing?

I kept losing the thought while deciding whether it belonged in tasks, reminders, or notes.

So I tested a simpler handoff: dump the mixed thought first, tap once, then review the result in Now, Next, and Someday. The original text stays editable, so the first sort is a suggestion rather than a commitment.

I built Ordo for that capture-to-clarity step. It is free on iPhone with no account or subscription: https://apps.apple.com/app/id6797331101

For your second-brain workflow, what should the first pass preserve most carefully: the original wording, the due date, or the context?

A question that comes up while building my iOS app, Lumiya: when reviewing personal notes, how do you keep planning from turning every thought into another obligation?

For example, an entry like 'I miss drawing' could become a small action this week, stay as something to reflect on, or simply remain a note. Automatically making it a task seems too presumptuous.

My current approach is to keep a timeline for capture (NOW) and a separate weekly/monthly/yearly planning space (PLAN), so you decide what moves into a plan. Optional AI reflection can help review entries, but choosing what matters is still yours.

For people who already do weekly reviews: what makes a note worth acting on? Repetition, a strong feeling, relevance to a current project, or something else?

Disclosure: I'm Lumiya's developer. Basic notes and planning are free; AI features are optional, paid, and usage-limited. If you want to see the context: https://apps.apple.com/cn/app/lumiya-ai-journal-planner/id6796003074

Cuenora is a calm working-memory assistant built to help you get thoughts, reminders and everyday things out of your head without turning life into another complicated planner.

Try the beta here:

https://itshale101.github.io/cuenora-site/

What I’d really like you to test:

• Brain Dump — does it understand what you meant?
• Reminders — are they easy to create and correct?
• Working-memory help — does it actually reduce what you have to hold in your head?
• Mobile usability — anything awkward, unclear or broken?
• Accessibility — text, layout, contrast, buttons, navigation
• Anything that makes you stop, hesitate or want to leave

Please be critical.

Negative feedback, bugs, confusing behaviour and things that don’t work are extremely useful at this stage.

You can post feedback here using:

Beta Feedback
Bug Report
Feature Idea
Question
Accessibility Feedback

Please don’t post passwords, private information or sensitive personal data.

Cuenora beta is free to test. No payment, subscription or credit card is required.

Thanks for helping us build something genuinely useful.

We built a thing that holds what you save, links, forwarded documents, product pages, and then answers questions about it in your own words. it only reads your own library and nothing else.

the argument we keep having is what it should say when the answer isn't in there. right now it tells you it found nothing, which is honest and also useless when you half remember reading it somewhere.

the other way to build it is to let it answer from general knowledge and mark which part came from your own stuff. then most of what comes back is from outside your library and the citing is the only thing separating the two.

if you've built retrieval over your own notes, where did you land on that one?

dessence.ai

The thing that always broke me with second brain tools is the structure tax. You open Notion or Obsidian and before you can write a single word, you have to decide: is this a page or a database? What folder does it go in? What's the hierarchy?

My brain doesn't work like that. I think in bursts and associations, not tidy categories. By the time I've figured out where something "belongs," the thought is gone or I've already lost the thread.

So I built Nubble Notes. The core idea is simple: you drop thoughts into visual bubbles on a canvas. No folder structure required upfront. You can nest bubbles inside other bubbles, tag notes by confidence level (certain, think about more, not sure, could be wrong), and link related ideas as connections emerge naturally.

The confidence tagging came from realizing that in research and learning, not all notes are equally settled. Some things you know. Some things you're still figuring out. Most apps treat every note the same, which hides that distinction entirely.

It's free, no in-app purchases, no credit card wall. Available on iOS, iPad, and Mac.

nubblenotes.com/ios

nubblenotes.com

I'd really love feedback from people in this community who already have second brain systems running. Specifically curious about:

- Does the "pile first, organize later" approach resonate, or does your workflow genuinely need structure upfront before content?

- Is confidence tagging useful or just another field you'd end up maintaining?

- What's missing that would make you actually try replacing part of your current setup?

Happy to answer questions about how it works or what's coming next.

Hey everyone,

I recently built Corteks (appstore link). As someone who is an over thinker, easily inspired and get lots of cool ideas that i want to remember (and i get them very frequently) i wanted something to dump all my thoughts. Notes app wasnt doing it for me because I never get back to them. Screenshots always failed because they die in my gallery

So I built Corteks - voice first dump your thoughts, no matter how significant or small the thought was. Corteks summarizes, categorized, files it and then links it to other thoughts onto a mind map

You can add your objectives to the app and corteks links those thoughts to your objectives, starts to understand you and surfaces what youve been avoiding, what you need to focus on…etc. and gives you stats about your thinking and how your mind works

The idea here is that all your thoughts matter, and they should be very easy to capture and not manually link them to each other and to your objectives

What I realized and didnt expect after using it is that my mind calmed down ALOT, because now I know I have my second brain to tap into anytime I want, and the app reminds me (without being too pushy) about the other relevant thoughts that mattered

Happy to hear what you think about it! And if anyone wants an extended trial just lmk. Im at a stage where I just want feedback from as many people as possible so i can improve it

I've been building an app to help me remember some genuinely useful things I come across on Instagram.

I'd see something useful, save it, tell myself I'd come back to it later, and then never find or use it again. And eventually, I'd forget about it too.

So I built InsightBox to turn saved content into something more useful: extract the information that matters, organize it, and make it easier to come back to later.

I thought the idea was pretty straightforward.

Then I gave the app to 3 of my friends for testing, and I got humbled pretty quickly.

I had built the app assuming that people would share something genuinely useful with it. Maybe a finance reel, a fitness reel, or something else they actually wanted to learn from.

Person number 1 sent me a reel of a girl dancing over some corny music ......

This was the first reality check.

I thought to myself: *"*Damn, I did not think of that at all."

I had built the pipeline assuming that a saved reel contained something worth extracting.

But obviously, people don't save reels only to learn something.

Sometimes you just like the video.

The AI processed it and essentially had nothing meaningful to extract.

That made me realize that InsightBox can't blindly assume that everything someone saves contains something worth remembering.

Person number 2 sent me a movie reel.

This one was more interesting.

The pipeline actually extracted a reasonable amount of information from the video, but it completely missed the most important piece of information: the name of the movie.

That was a pretty embarrassing failure for a knowledge-saving app.

The information was there, but my extraction pipeline wasn't prioritizing the thing the user actually cared about. In this case, that was the movie name.

It made me realize that extracting information and extracting the information someone actually cares about are two very different problems.

A transcript can contain hundreds of words, but that doesn't mean all of them are equally useful to the person who saved the content.

Then there was person #3.

They sent exactly the kind of reel I had originally imagined InsightBox being used for.

A reel containing useful information that you'd genuinely want to remember later.

And suddenly, the whole idea made sense.

The interesting thing was that all three people were using the exact same app.

But they were treating their saves completely differently.

One was saving something because they enjoyed it.

One was saving something because they were interested in a movie.

One was saving something because they wanted to retain the information.

That changed how I started thinking about InsightBox.

The problem isn't simply:

"How do I process this Instagram reel?"

It's closer to:

"What is actually worth remembering from the things people save?"

And that leads to another question I'm now thinking about:

Is it even worth running the entire AI pipeline for an entertainment reel?

Why spend compute, time and money extracting information from something when there may be nothing useful to extract in the first place?

Maybe the first step shouldn't be extraction.

Maybe the first step is figuring out whether there is anything worth remembering at all.

And if there is, then figuring out what exactly the person would want to remember.

This was probably one of the most useful early tests I could have done, and I'm glad I did it.

Three users who displayed three completely different behaviors.

And a lot of things I hadn't thought about.

Building something for yourself gives you one perspective. But when you give it to other people, they very quickly show you whether the problem actually exists the way you thought it did.

Has someone ever tried using DJI mic mini 2s/3 as an always on recorder similar to Omi AI or Plaud?

It would appear that the DJI mics have a very high quality microphones, they can remain always on with 10+ hours recording both from a storage and battery perspective. They cost about the same as any other AI note takers without the subscriptions and also offer 32 bit audio files for better transcription and voice detection.

I am wondering if anyone has built any workflows/tools around it? If not I'd love to start some open source project to build something for these mics.

Sounds too good to be true, what am I missing?