athena-oracle/transcripts/ep_obsidian_omi.clean.txt

# Agent OS + Obsidian + Omi Is INSANE! — AI News Today | Julian Goldie Podcast
# Transcript (cleaned) — 8:27 episode
# Source: https://open.spotify.com/episode/47Emp3Rnz0DQLj0FfRlHTl
# Resolved RSS: https://feeds.transistor.fm/ai-news-today-julian-goldie-podcast
# MP3: https://media.transistor.fm/51d728ef/8613bf18.mp3
# ASR: faster-whisper tiny/int8/beam1 (CPU). Proper nouns restored from raw.

[0:00] This is a powerful way to have a memory system, as you can see right here. That just works, and it has a powerful set of context, and basically this trains all of our agents on exactly what we've been working on. So that when we plug them all together, for example, all of our own agents — ChatGPT, Claude — understand Hermes. All of our agents work together, and they never forget context again.

[0:28] Now, at the same time, we also have NotebookLM. NotebookLM is one of the most powerful tools that I've ever seen for creating research. And also the amazing thing about NotebookLM is, like, you can generate, for example, videos in a single click for free using this system as well. So for example here, you can see all of these videos that we have built, and these are free videos. We can generate fully educational, fully set-up, fully automated videos in one single click, and we can use NotebookLM to do that.

[1:01] Now when you combine the two — both of these free tools, NotebookLM, and then Obsidian as well — you get a powerful second brain that can basically build and automate anything together. So you might be wondering, okay, why would you want to use something like Obsidian? Well, for example, if I go into Claude here and I'm like, "what did I work on yesterday?" — it's not going to have much of an idea of what I'm doing, what I'm working on, what's useful, what's not useful, etc. And so what you want to do instead is have all of your agents in one shared memory system, as you can see — where you've got Hermes, you've got Claude, you've got Antigravity, you've got, for example, Codex — and they all operate from the same shared memory. That means all of your agents understand each other.

[1:38] You might say, why would you do that? Well, I call this the Infinite Context Engine. It's one memory that every agent shares, and your chat trains your vault, your vault trains your agent, and this is a loop that gets smarter forever. So all of your memories link to, for example, all of your goals, your projects, your areas, your daily notes — and it just links beautifully.

[1:55] The problem with this, if you use something like ChatGPT, is: imagine if you hired an assistant every morning, they wake up, they remember nothing about you, you explain your whole life again — your name, your business, your goals, your clients — every day. Well, that's essentially what your agents are like right now, because they don't have a memory system; your AI is genius, but it has amnesia. Every chat you start from zero. The Infinite Context Engine that I'm showing you today breaks this cycle for good.

[2:23] You might say, well, AI memory sounds complicated and technical. All this is, is a bunch of Markdown files organized into a nice system. That is literally it. This is just a bunch of files and folders that all link together naturally to train your agents on all of your context. And you don't even need coding to do this. You can get your agents to automatically update it for you, and you can get your agents to automatically come up with the ideas for it.

[2:55] Now, there's three tools with one shared brain that I use to set this up. So we have Omi, we have Obsidian, and we have the AI. Omi is basically a system where we can automatically record our screen, our microphone — it's looking at my device all the time, taking an example and looking through all of what we're doing day to day. And then you can see here, it takes notes on what I'm doing, what I'm working on, and it does this like every hour.

[3:17] When we're doing this, this actually exports into Obsidian. Obsidian is where we have our knowledge graph — everything linked together, all of our Markdown files. So what you end up with is Omi creating the memories and taking the memories automatically for you, and then that plugs into Obsidian. And Obsidian is where all your agents come together. If we click on this, you can see it has all of the information about this particular topic, neatly organized, formatted nicely, everything links together. It's just like a Wikipedia, but for agents — everything links together, and you've got everything in one place.

[3:55] So we've got Omi taking the notes. You've got Obsidian gathering the notes that you've created and organizing them. And then that plugs into an agent operating system — which is where you have Claude, you have Omi, every CLI all pointed at the same vault. So every one of them wakes up already knowing you, with one memory in every agent.

[4:16] You might say, well, this works with one tool but not all of mine. But the vault is just text files. Any AI that's trained to read a file path can read your memory. So it's one setup — Omi, every tool with the same brain working together. That's how it works: you capture with Omi, you organize with Obsidian, you store it in Obsidian, and you deploy it to your agent operating system.

[4:38] And this is infinite, because most memory setups just work on what you feed the AI — it reads, and that stops improving the day you stop typing. The Infinite Context Engine is a loop. So every chat your agents have gets written back into Obsidian — automatically, one file per day. So the vault doesn't just store what you tell it; it stores what your agents do. Then the next agent reads that and it gets better and logs more. And then the vault gets richer — your work trains your vault, your vault trains your agents — they go round and round, smarter every time. As a loop.

[5:08] So this is a really powerful way to just dramatically improve your systems every single day. If we go into one of our agents and we ask it for personalized ideas — like, for example, Hermes — it's going to understand exactly what we've been working on recently, as you can see here, and it's going to have personalized recommendations based on what we're going to do, and that all comes from the memory system. Then that links to, for example, free Claude Code to call into Hermes, and everything else that we have inside this system.

[5:29] It's your whole mind as a galaxy. If you look at this system here, every memory is a star, every star links together. Mine has about 364 stars and 1,200 links right now. And every link between notes is a line of light. The stars you've touched most recently glow the brightest. It's not just pretty — it's actually how you find things. So if you need a memory from three weeks ago, you don't dig through folders; you search a galaxy or follow the links from one star to the next. Your agent is doing exactly the same thing.

[5:58] When they're seen, this is what organized really looks like. You might say, well, my notes are a mess, this won't help. But that's the point — the galaxy connects the mess for you. Wiki links pull related notes together so you can see the threads. Omi, your chats, your notes — and then that goes out to Hermes, Claude, Open Claude, and every other CLI they use.

[6:21] If you think about this: the old way, without a memory, is like you lose loads of times because you have to explain who you are every session. You pay the same context in every tool. You get generic answers that don't know your business. Each AI tool starts from scratch, and you forget what you decided last week. With the new way — with the Infinite Context Engine — it just knows you. Every agent wakes up already knowing you, one brain shared by a memory with Hermes. It gives answers built on your real goals and clients. It finds any past memory in seconds. The galaxy and the loop make this better every day.

[6:45] The other thing we built in over here is search. We can type in a memory like this, and it will search across all of our memories and pull up the latest notes on that. So you'll think, like, "Julian's meaning in life" — this is pretty amazing; it's actually understood what I see as meaning from life, and then it's detailed that into beautiful documentation.

[7:02] So that's how the whole system works — Obsidian, and then that links to all your AI agents. Really powerful stuff. You can see how it gets you back. You might say it's complicated — it's not, because it's really simple and easy to just implement. You might say this is technical — but you get your agents to do it for you. You might also say, well, does this use a lot of tokens? But we actually have token-minimization playbooks if you're worried about all that sort of stuff inside the app.

[7:25] So if you want to get the Infinite Context Engine built for you, you can wire this up yourself, or you can get the whole thing done inside the agent operating system — inside their platform, with the full Obsidian and memory-galaxy setup already wired, the Omi and Obsidian setup walkthrough, step-by-step coaching calls where I set up the memory with you, the community of 4000 operators running their six-hour stack, plus the prompts and a member map for your city.

[7:44] So if you want to get all of that, I've got a profit link in the comments description. Also get the full agent OS — if you want to get this, it's inside the classroom; go to "new daily updates" and you'll find the Agent OS over here, with the zip file to install it and a video tutorial. We also have daily tutorials on how the base actually works and what's useful. And inside the community, I answer all the questions, with a video tutorial every single day. Inside the calendar you can jump on coach calls, ask questions, wire this setup together, and meet up with cool people doing the same thing. Inside the map, you can meet people locally near you who are building with AI agents just like you.

[8:18] So feel free to get it — link in the comments description, or go to the app. Bye — so watch it.

---
NOTE: Light ASR (tiny) leaves a few ambiguous terms. Best-guess restorations:
- "blog.com" (end) → likely the AI Profit Boardroom / Agent OS app URL (unclear in audio).
- "Open Hall" / "Open Claude" → treated as Claude references.
- "room of 4000 operators" → "community of 4000 operators".
Raw file (verbatim ASR): ep_obsidian_omi.txt (same dir).