# Give Your AI Unlimited Updated Context
Canonical: https://social-archive.org/adegette/CcVAUZkKhS
Original URL: https://saranfn.substack.com/p/give-your-ai-unlimited-updated-context
Author: Sara Nóbrega
Platform: substack
## Content
• • TLDR: Every time you open a new AI conversation, your context resets. Built-in memory features help a little, but they don’t know about your active projects, your team, or what happened yesterday. The solution is a local folder of plain text files that any AI can read, called a vault. I built one using Obsidian, and it changed how AI actually works for me on a daily basis. This article explains the problem and the concept. Part 2 is the full step-by-step setup (stay tuned!). The goal: One folder. One vault. One context to rule them all. Ok, enough of LOTR references already. Maybe the most important insight you can get from this article is not why context itself is important, but why you should care about portability. Why is this important? If tomorrow you/your team decide to switch from ChatGPT to Claude, or to Claude from Gemini, you have a fast and efficient way to transport your up-to-date context. The probability that you will somewhere in time switch to another AI is actually high. I will leave this topic to another article! • • Every AI Conversation Starts Blank. That’s a Real Problem.You’ve probably felt this. You open Claude, you start typing, and within two sentences you’re re-explaining who you are, what your business does, what you’re working on, and what you decided last week. Again. It’s not just annoying. It’s a structural flaw in how most people use AI right now, and it’s costing real time every single day. I started noticing this pattern with my own workflows and with people I help set up AI systems. The tool isn’t the bottleneck. The missing context is! Before we start: • Save this and spend 20 minutes today thinking about how to improve your context/memory layer. It will be worth it! • Send it to any engineer, founder, or friend who complains about re-explaining themselves to AI every day. Share Why Built-In Memory Isn’t EnoughClaude and ChatGPT both have memory features. They remember things like your name, your job title, and maybe a few preferences. That’s useful. But it’s not the kind of memory that actually matters for work. What memory features don’t capture: your active projects, the decisions you made last month, who’s on your team and what they own, what happened in your project management tool this week, the deal you’re about to close, or the vendor you just ruled out. • • Think of it like this: built-in memory is like a sticky note on the fridge. It tells you someone likes oat milk. It doesn’t tell you there’s a dinner party on Saturday, three guests are vegetarian, and you already bought the ingredients. For real work, you need the full kitchen. The deeper problem is portability. • • Even if you get one AI model working well with your context, that context lives inside that tool. Switch to a different AI, open a new interface, or start a fresh session and you’re back to square one. The Concept: A Vault That Any AI Can ReadHere’s the shift I’ve been building toward: What if your context didn’t live inside any AI? What if it lived in a folder on your computer, in plain text, that any AI could read? That’s the idea behind the vault approach. You keep all of your business context, your projects, your team, your decisions, your current priorities, in a local folder of Markdown files. When an AI needs context, it reads the folder. When something changes, the folder updates. The AI is just the reader. The folder is the memory. • • This is different from RAG pipelines or vector databases. Those are engineering solutions that require real setup, maintenance, and technical skill. A vault is just files. Text files in folders. Any AI that can access your file system can read them. Markdown is the format LLMs were trained on. There’s no translation layer needed. When Claude reads a .md file, it’s reading the same format it already understands natively. That’s what makes this surprisingly effective with almost zero overhead. Learn AI is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. Why ObsidianI’ve been using Obsidian as the front end for this. It’s a free note-taking app that stores everything as plain .md files in a regular folder on your Mac or PC. No cloud account, no subscription, no proprietary format. • • If Obsidian disappeared tomorrow, you’d still have every file! Your data lives on your hard drive, not in someone else’s database. If you’ve ever been burned by a SaaS tool shutting down or changing its API, you understand why that matters. But the real reason I chose Obsidian isn’t about Obsidian itself. It’s about the folder. Obsidian is just a visual layer on top of a regular directory. Claude reads the same folder. They’re looking at the same files on disk. No syncing, no API, no middleware between them. What Goes in the VaultThe contents depend on your work, but the categories tend to be consistent across different people and businesses. Here’s what I’ve found works: • • • A business map: what your company does, all your entities, locations, revenue streams • Your team: who’s who, their roles, and enough context that an AI can reference them correctly • Active projects: one file per project, with current status and next actions • Decisions log: choices you’ve made, tools you’ve ruled out, strategic calls • Preferences: how you like to communicate, what formats you prefer, what tone to use • A daily digest: pulled automatically from your tools every morning The key file is CLAUDE.md at the vault root. This is an instruction file that tells any AI how the vault is structured and where to find things. Claude reads it automatically at the start of every session. It’s the table of contents for your context. The Part That Makes It Actually Useful: Auto-UpdatingA vault that requires manual updates is a vault you’ll stop maintaining within two weeks. The version I’ve been running uses Claude’s Cowork mode to run a scheduled task every weekday morning. It pulls from my data sources and summarizes what happened in the last 24 hours, updates the relevant project files, and appends a dated entry to a daily folder. The vault stays current without anyone touching it. This is the workflow chain: • • When you then open a conversation with Claude, it reads the CLAUDE.md file and knows exactly where to find current context. No re-explaining. No setup. Just work. What This Unlocks in PracticeHere are a few real examples of how this changes day-to-day AI use: Status check: You ask “what’s the status of the Marketing project?” -> Claude reads the Roswell project file -> gives you a current update without you touching anything Follow-up triage: You ask “who do I need to follow up with today?” -> Claude reads the daily digest and contacts file -> gives you a prioritized list based on what actually happened yesterday New AI, same context: You want to try a different AI tool -> you point it at the same vault folder -> it reads the same files -> your context travels with you Onboarding context fast: You’re jumping into a new kind of task -> you open a fresh session -> Claude already knows your business, your team, your open threads Personal TakeI want to be clear about what this is and what it isn’t. It’s not magic. The vault is only as good as what’s in it. If you never update the project files, Claude reads stale context. If your CLAUDE.md file is vague, Claude won’t know where to look. The system needs some upfront investment to set up properly, and it needs the automated update layer to stay useful over time. What surprised me: the impact on how you use AI changes more than I expected. When the context is already there, you stop using AI for one-off questions and start using it for actual work. The conversations feel completely different. I set this up for my own workflows and I also helped someone I work with build it out for their business. In both cases, the biggest shift wasn’t the tool. It was the realization that the re-explaining problem was solvable, and that the solution was simpler than expected. Why This Makes You a Power User of AIMost people use AI reactively. They open a chat, they explain their situation, they get a response, they move on. Every session is isolated. Every conversation starts from zero. A vault changes that completely. You’re not just using AI. You’re running AI with persistent, portable, structured context. That’s a qualitatively different thing. The competitive advantage is real. If you’re a founder, operator, or builder working with AI every day, the people who have solved the context problem are getting dramatically more out of the same tools. They’re not re-explaining. They’re not losing decisions to session resets. They’re compounding. And because the vault is just files, it works with any AI. You’re not locked into Claude. You’re not locked into any tool. Your context travels with you as the landscape changes. What’s NextIn Part 2, I’m going to walk through the exact setup I used. • • That includes how to extract all the context that already exists in your AI conversations, how to design a folder structure from your actual tools instead of a template, how to write the CLAUDE.md instruction file, and how to wire up the daily automated update. It takes a few hours to set up. After that, you don’t touch it. If that sounds worth it, stay subscribed! 🎯 Need help navigating your own AI journey? Right now I’m offering: • AI consulting calls - If you’re working on projects or exploring how to bring AI into your business, I’d love to jump on a call and help you figure things out. • Mentorship sessions - Thinking about a career move or growing your skills in AI? I can share advice and help you map out your next steps. • Premium membership perk - If you’re a premium subscriber to this newsletter, you already have two 1:1 sessions with me included. • AI Engineering Mentorship Project - A structured, hands-on program where we build and deploy an AI/MLOps project together. You’ll learn production-grade skills through real code and weekly guidance. For more info DM me or reply to this email!. Learn AI is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber. --- TLDR:
Every time you open a new AI conversation, your context resets. Built-in memory features help a little, but they don’t know about your active projects, your team, or what happened yesterday. The solution is a local folder of plain text files that any AI can read, called a vault. I built one using Obsidian, and it changed how AI actually works for me on a daily basis. This article explains the problem and the concept. Part 2 is the full step-by-step setup (stay tuned!).
The goal: One folder. One vault. One context to rule them all. Ok, enough of LOTR references already.
Maybe the most important insight you can get from this article is not why context itself is important, but why you should care about portability.
Why is this important?
If tomorrow you/your team decide to switch from ChatGPT to Claude, or to Claude from Gemini, you have a fast and efficient way to transport your up-to-date context.
The probability that you will somewhere in time switch to another AI is actually high. I will leave this topic to another article!
Every AI Conversation Starts Blank. That’s a Real Problem.
You’ve probably felt this.
You open Claude, you start typing, and within two sentences you’re re-explaining who you are, what your business does, what you’re working on, and what you decided last week. Again.
It’s not just annoying. It’s a structural flaw in how most people use AI right now, and it’s costing real time every single day.
I started noticing this pattern with my own workflows and with people I help set up AI systems.
The tool isn’t the bottleneck. The missing context is!
Before we start:
Save this and spend 20 minutes today thinking about how to improve your context/memory layer. It will be worth it!
Send it to any engineer, founder, or friend who complains about re-explaining themselves to AI every day.
Share
Why Built-In Memory Isn’t Enough
Claude and ChatGPT both have memory features.
They remember things like your name, your job title, and maybe a few preferences. That’s useful. But it’s not the kind of memory that actually matters for work.
What memory features don’t capture: your active projects, the decisions you made last month, who’s on your team and what they own, what happened in your project management tool this week, the deal you’re about to close, or the vendor you just ruled out.
Think of it like this: built-in memory is like a sticky note on the fridge. It tells you someone likes oat milk. It doesn’t tell you there’s a dinner party on Saturday, three guests are vegetarian, and you already bought the ingredients. For real work, you need the full kitchen.
The deeper problem is portability.
Even if you get one AI model working well with your context, that context lives inside that tool. Switch to a different AI, open a new interface, or start a fresh session and you’re back to square one.
The Concept: A Vault That Any AI Can Read
Here’s the shift I’ve been building toward:
What if your context didn’t live inside any AI? What if it lived in a folder on your computer, in plain text, that any AI could read?
That’s the idea behind the vault approach.
You keep all of your business context, your projects, your team, your decisions, your current priorities, in a local folder of Markdown files.
When an AI needs context, it reads the folder. When something changes, the folder updates. The AI is just the reader. The folder is the memory.
This is different from RAG pipelines or vector databases. Those are engineering solutions that require real setup, maintenance, and technical skill. A vault is just files. Text files in folders. Any AI that can access your file system can read them.
Markdown is the format LLMs were trained on.
There’s no translation layer needed. When Claude reads a .md file, it’s reading the same format it already understands natively. That’s what makes this surprisingly effective with almost zero overhead.
Why Obsidian
I’ve been using Obsidian as the front end for this. It’s a free note-taking app that stores everything as plain .md files in a regular folder on your Mac or PC. No cloud account, no subscription, no proprietary format.
If Obsidian disappeared tomorrow, you’d still have every file!
Your data lives on your hard drive, not in someone else’s database. If you’ve ever been burned by a SaaS tool shutting down or changing its API, you understand why that matters.
But the real reason I chose Obsidian isn’t about Obsidian itself. It’s about the folder. Obsidian is just a visual layer on top of a regular directory. Claude reads the same folder.
They’re looking at the same files on disk. No syncing, no API, no middleware between them.
What Goes in the Vault
The contents depend on your work, but the categories tend to be consistent across different people and businesses. Here’s what I’ve found works:
A business map: what your company does, all your entities, locations, revenue streams
Your team: who’s who, their roles, and enough context that an AI can reference them correctly
Active projects: one file per project, with current status and next actions
Decisions log: choices you’ve made, tools you’ve ruled out, strategic calls
Preferences: how you like to communicate, what formats you prefer, what tone to use
A daily digest: pulled automatically from your tools every morning
The key file is CLAUDE.md at the vault root. This is an instruction file that tells any AI how the vault is structured and where to find things.
Claude reads it automatically at the start of every session. It’s the table of contents for your context.
The Part That Makes It Actually Useful: Auto-Updating
A vault that requires manual updates is a vault you’ll stop maintaining within two weeks.
The version I’ve been running uses Claude’s Cowork mode to run a scheduled task every weekday morning. It pulls from my data sources and summarizes what happened in the last 24 hours, updates the relevant project files, and appends a dated entry to a daily folder. The vault stays current without anyone touching it.
This is the workflow chain:
When you then open a conversation with Claude, it reads the CLAUDE.md file and knows exactly where to find current context. No re-explaining. No setup. Just work.
What This Unlocks in Practice
Here are a few real examples of how this changes day-to-day AI use:
Status check: You ask “what’s the status of the Marketing project?” -> Claude reads the Roswell project file -> gives you a current update without you touching anything
Follow-up triage: You ask “who do I need to follow up with today?” -> Claude reads the daily digest and contacts file -> gives you a prioritized list based on what actually happened yesterday
New AI, same context: You want to try a different AI tool -> you point it at the same vault folder -> it reads the same files -> your context travels with you
Onboarding context fast: You’re jumping into a new kind of task -> you open a fresh session -> Claude already knows your business, your team, your open threads
Personal Take
I want to be clear about what this is and what it isn’t.
It’s not magic. The vault is only as good as what’s in it. If you never update the project files, Claude reads stale context. If your CLAUDE.md file is vague, Claude won’t know where to look. The system needs some upfront investment to set up properly, and it needs the automated update layer to stay useful over time.
What surprised me: the impact on how you use AI changes more than I expected.
When the context is already there, you stop using AI for one-off questions and start using it for actual work. The conversations feel completely different.
I set this up for my own workflows and I also helped someone I work with build it out for their business. In both cases, the biggest shift wasn’t the tool. It was the realization that the re-explaining problem was solvable, and that the solution was simpler than expected.
Why This Makes You a Power User of AI
Most people use AI reactively.
They open a chat, they explain their situation, they get a response, they move on. Every session is isolated. Every conversation starts from zero.
A vault changes that completely. You’re not just using AI. You’re running AI with persistent, portable, structured context. That’s a qualitatively different thing.
The competitive advantage is real.
If you’re a founder, operator, or builder working with AI every day, the people who have solved the context problem are getting dramatically more out of the same tools. They’re not re-explaining. They’re not losing decisions to session resets. They’re compounding.
And because the vault is just files, it works with any AI. You’re not locked into Claude. You’re not locked into any tool. Your context travels with you as the landscape changes.
What’s Next
In Part 2, I’m going to walk through the exact setup I used.
That includes how to extract all the context that already exists in your AI conversations, how to design a folder structure from your actual tools instead of a template, how to write the CLAUDE.md instruction file, and how to wire up the daily automated update.
It takes a few hours to set up. After that, you don’t touch it.
If that sounds worth it, stay subscribed!
🎯 Need help navigating your own AI journey? Right now I’m offering:
AI consulting calls - If you’re working on projects or exploring how to bring AI into your business, I’d love to jump on a call and help you figure things out.
Mentorship sessions - Thinking about a career move or growing your skills in AI? I can share advice and help you map out your next steps.
Premium membership perk - If you’re a premium subscriber to this newsletter, you already have two 1:1 sessions with me included.
AI Engineering Mentorship Project - A structured, hands-on program where we build and deploy an AI/MLOps project together. You’ll learn production-grade skills through real code and weekly guidance.
For more info DM me or reply to this email!.
