Everyone knows that enthropic models are incredible, especially when it comes to coding. But when you're working with large codebases and long context, you'll sometimes notice a problem. Even when the model is quite powerful itself, like the Opus 4.6, they can still lose focus over time and occasionally forget steps you previously gave them. And when you're building a complex project, that can slow things down. But today, I want to showcase a simple solution, and it's with a tool called Obsidian, which is completely free. Obsidian is a notetaking app that stores everything as a plain markdown file inside your folder on your computer. This is called the vault. This means your notes are not just text files; there's no proprietary format, no lockups, it's fully private and completely offline. But what makes it perfect for Claude Code is that it keeps your memory or your context persistent. It keeps it structured, and it will be there for your entire project, for the architecture decisions, the design choices, key instructions, session summaries, to-do lists, bug notes, naming conversations, basically anything that matters. Because when it comes to readable markdown files in a folder, Claude Code can directly read from the vault. It can search it, it can reference specific notes, and it can even update them so that it could build on sessions after sessions. This gives Claude the long memory that it normally lacks in a chat interface or simply through the Claude Code instance. So, you no longer have to reexplain the same context every single time. The model will stay aligned with your original vision. It will remember important details across weeks or even months, and it produces far more consistent and focused code. This will overall dramatically reduce the forgetfulness and the problem that often happens in large code bases.
So, in short, when you combine Obsidian with Claude Code, you have persistent memory, and it enables a second brain for your codebase. This will overall get you better generations, smoother workflows, and far less frustration when you're working on large, long-running projects. Still running object detection with NMS pipelines in 2026? Well, it's time for you to upgrade. And this is where I would like to introduce Ultralytics YOLO 26, the next-gen edge-first vision model built end-to-end from the team behind YOLO version 5, YOLO version 8, as well as YOLO 11. YOLO 26 removes the non-maximum suppression entirely.
This means true NMS-free end-to-end inference, no fragile post-processing, lower latency, as well as cleaner deployment. It removes distribution focal loss for better large object reliability and hardware compatibility. It also adds stall enhancements for stronger small object detection. It's up to 43% faster with CPU inference in real-time on existing hardware and built for edge devices in production systems. Plus, the new MUSGD optimizer improves training stability and convergence. This isn't built to win benchmarks; it's built to win in production. It's open source as well under the AGPL 3.0 for research, and enterprise licensing is available for commercial deployment through Ultralytics. So, if you're looking to try this out, you can do so because YOLO 26 is live now. You can try it out on the Ultralytics platform, which is today's video sponsor. Built end-to-end, built for edge, and this is the new standard for real-time vision AI.
To get started, just make sure you have Claude Code ready and installed on the latest update, as well as having Obsidian installed, which is completely free to install. I'll leave a link to this in the description below. Now, after installing, you will have this popup, and this is where you're going to need to create your vault. Now, what is a vault? Well, Obsidian basically stores all of your notes within your specific vault in a plain markdown file. This means everything is just a normal text file, but it will be readable, private, and easy for our tool like Claude Code, for example, to easily access. Because these files live within this particular folder, Claude can see them, read them, search them, and update them during your coding sessions. Over time, the vault will become the persistent knowledge base for your project, storing all of the important things that we're talking about that are prevalent within any codebase, like the architecture decisions, the coding rules that you have set, bugs, session summaries. All of this can be updated and present within this one folder so that in future sessions, it doesn't lose any sort of memory or any sort of behavior that you had taught it. So, give it a name, and you can select the location where you want to place this vault.
We have created the vault, but what do we put inside it? Well, this is where you want to simply start off with your file structures. You want to make sure that you are putting in all the relevant context that your Claude Code agents can use. Make sure you provide your GitHub-related information, whether that's your codebase, PRDs, research, anything that Claude can reference when making decisions. You can even provide meeting notes, like transcripts that Claude can scan to recall past conversations from different sessions. You can even provide personal notes, ideas, writing, and long-term thinking that you personally have. So, we have created the vault, and I have now added in a couple of different files, a few simple files like the project overview and my daily notes.
This is essentially where you want to provide as much context as possible, whether that's coding rules or session logs. Think of this as a memory file system for your AI. Instead of trying to remember everything inside the chat within the Claude Code instance, Claude can read these notes whenever it needs them because they are in an MD format. This is where it gets powerful because I can simply reference certain files or certain text stacks within this Obsidian vault before generating the code so that Claude can check how the project is structured, what frameworks I'm actually using, and the coding conversations. What I really like about Obsidian is that it gives you a graph view of everything as well, so you get a better contextual idea of your real project memory. Now, what I would also recommend is installing the Agent Skills to use Obsidian better because this will teach Claude Code how to interact with your vault more effectively. It has predefined skills that will enable it to read your markdown files easier. So, Claude can parse your notes or any relevant information better. It can also write and update notes for you, which is great because after a Claude session has completed, you can have it write up a session note and upload it directly to your vault, like I did over here. After coding out, I created this simple note after that session so that whenever I pick up another Claude Code session, I can reference this previous note so that it has that persistent memory in future generations. It also has the ability to search and reference content, maintain persistent memory through a normal chat instance, as well as working with structured data.
To get started, copy this first command, which is to install the marketplace for this plugin. You can then add that marketplace and then run the plugin install command to set up the Obsidian skill. You can set this up for your instance. Once that is ready, you can get started with your codebase. Now, to showcase our second brain in action and what it's capable of doing, we're going to have it first reference a prior project. This is where I had an active project, a CRM dashboard with this text stack that I had built previously. Now, there are a lot of missing features. We're going to have Claude Code first pull the project context. We're going to have it build a deals Kanban board using the Prisma deal model. We're going to have it follow the CRM component architecture, as well as the Shad CN UI patterns that were previously active within our context. This is where it will first start off by checking out the component architecture, knowing which Shad CN components you actually use, also referencing reusable patterns by reading the Shad CN UI patterns for forms, tables, modals, and toast notifications, and lastly, following the coding standards. And it can even update the vault because after it has finished generating this, it will update it based off the context, the new context that we had generated, so that for future generations, it can reference this session to get better, smarter results even further in the future.
You can see that right now it is referencing the existing CRM project and the Prisma schema and it's able to reference the detailed information that we have present within our vault. So rather than having it just find context every single time, this persistent memory will be stable and always there within this vault so that Claude can continuously use it whenever it needs it. As an FYI, when you're working with the new Obsidian skill that we had installed, you can simply type /obsidian and then open up the Obsidian CLI so that Claude can use the Obsidian CLI to either read, create, or edit notes. It can search the vault. So, if you want to ingest prior context, you can use the search vault command so that it gives your Claude Code instance the particular knowledge that it needs. You can manage tasks, daily notes, as well as plugin development. When you're working on a larger project with a lot of sub-agents, it will be able to clearly describe the context across multiple agents so that there's no deviation from what you're working on. All the agents will have the same mindset, and they will keep the same context as they work through the progress. In this case, one agent could be working with the deal pipeline, and the other one could work on the analytics chart. Over time, it will keep the same context, the same structure, the same text stack as it develops through everything. Thanks to Claude Code getting the correct context that it needed for building out a new feature for our CRM dashboard, like the Shad CN UI components, it was able to build out the Kanban board feature that we had requested.
This is where it was able to keep the consistency of our overall dashboard. It was able to use the correct components, and you can see that it coded out this beautiful functionality with animations and added in all of the components that we had looked for. That is what this new second brain solves; it makes your life easier when you're working with these AI agents. What's great is that the vault actually grows too because Claude always has up-to-date project memory when you reference the vault. After you have gotten the context, you can even generate and update the code and provide daily notes on what was accomplished so that your overall notes will be able to stay up to date. For example, you can simply use the Obsidian skill, the markdown skill, so that it is able to create a new note, providing a new daily note that our future session can revisit.
That way, it is able to get the correct context as to what was added or built. You can see right now that it is searching our previous pattern of the CRM dashboard that I built, and it's going to create a new note on whatever it had developed. If you like this video and would love to support the channel, you can consider donating to my channel through the Super Thanks option below. Or you can consider joining our private Discord, where you can access multiple subscriptions to different AI tools for free on a monthly basis, plus daily AI news and exclusive content, plus a lot more. But that's basically it, guys. I hope you enjoyed today's video. This is essentially where you can have Claude always have up-to-date project memory with this second brain. All you have to do is make sure that you have the correct structure set up with your file structure, and I will leave this article in the description below. It's not extraordinary or anything, but it tells you the basic structure that you need when you're working with this workflow. But that's basically it, guys.
I hope you found this video to be helpful. This will ensure your AI sub-agents within Claude have the correct context that they need over time. I'll leave all these links in the description below. With that thought, guys, thank you so much for watching. Make sure you join the newsletter, subscribe to our second channel, join our Discord, follow me on Twitter, and lastly, make sure you subscribe, turn on the notification bell, like this video, and please take a look at our previous videos so that you can stay up to date with the latest AI news. But with that thought, guys, have an amazing day, spread positivity, and I'll see you guys fairly shortly.